Nepal’s story reveals a fact: digital space has become part of national territory. If a government cannot establish effective sovereignty in digital space, its capacity to govern in the physical world will also disintegrate.

The cloud sounds weightless, dematerialised, floating above the planet. Tricontinental’s dossier no. 46, Big Tech and the Current Challenges Facing the Class Struggle writes plainly: ‘A data “cloud” sounds like an ethereal, magical place. It is, in reality, anything but that.’ The cloud is a set of extremely concrete, highly centralised infrastructure: server farms, submarine cables, chip fabrication plants, cooling towers – overwhelmingly located on US soil, subject to US law and US corporate control.

For most countries of the Global South, digital sovereignty remains a distant concept. They have not even begun to examine their own situation.

The Cloud Has a Physical Address

What does the landscape of global digital infrastructure look like?

Eighty per cent of Africa’s internet traffic travels through submarine cables owned and operated by European and US companies. The consequences of that dependence became visible in March 2024, when an underwater landslide off the coast of Côte d’Ivoire severed four cables simultaneously. Thirteen African countries along the western seaboard, including Ghana, Nigeria, and Côte d’Ivoire, lost internet access for weeks. Millions of users were cut off. The outage cost Nigeria alone over $590 million in four days. The companies that own the cables assumed no responsibility for the damage to African economies; repair timelines were determined by the cable owners, not by the countries affected. Afterwards, SpaceX’s Starlink rapidly expanded its market share in the region. Once dependency is established, crisis only deepens dependency.

As Bappa Sinha writes in Tricontinental’s Breaking the Stranglehold: How China is Shattering US Technological Hegemony, for over a century, the foundation of imperial power has been monopolistic control over the most advanced means of production, with military force and financial dominance as supporting instruments. From the industrial revolution through the age of digital platforms, successive imperial cores secured global dominance by capturing technological frontiers, extracting monopoly rents, and reinvesting those rents into further technological leadership, sustaining what appeared to be a self-reproducing hierarchy. The imperial core monopolises technology and capital; the periphery supplies labour and resources; unequal exchange continuously transfers value upward. This structure operates without formal colonial rule. When the Global South passed the New International Economic Order resolution at the United Nations in 1974, demanding technology transfer from North to South as a central provision, the response came two decades later through the Uruguay Round of the GATT: reverse engineering and technology transfer were made illegal. As Vijay Prashad writes in The Poorer Nations (Verso, 2012), what replaced the South’s demand was not a New International Economic Order but a North-led New International Property Order.

That order now governs the most advanced means of production of our own era: digital infrastructure. Chip manufacturing is concentrated in Taiwan and South Korea (using US-designed architectures). Operating systems are monopolised by Microsoft, Apple, and Google. The cloud computing market is dominated by Amazon AWS, Microsoft Azure, and Google Cloud. The global search engine market belongs almost entirely to Google. Countries, businesses, and citizens of the Global South use this infrastructure every day, but its physical location, jurisdiction, and control rest elsewhere.

Data is at the centre of all this. In April 2020, China’s State Council formally designated data as a fifth factor of production — alongside land, labour, capital, and technology. But the economic status of data is deeply ambiguous.

Current international accounting standards do not include data assets on corporate balance sheets. US tech giants, through their global platforms, continuously absorb data generated by users worldwide. No universally accepted method for valuing this data exists; what cannot be measured does not appear on any tax return. Facebook, Google, and Microsoft together avoided $2.8 billion in taxes across twenty developing countries in 2019 alone — a figure researchers describe as ‘the tip of the iceberg’. What is not measured is not taxed. What is not taxed is free.

This ‘unmeasurability’ is itself a mechanism of control. Global South countries cannot even quantify how much value they are losing, let alone assert rights over their data.

Three Structural Challenges, One Structural Trap

The Global South’s digital sovereignty predicament can be decomposed into three structural challenges. They reinforce each other and together constitute a structural trap.

1. Severe external dependence on digital infrastructure

From hardware to software to information security, Global South countries rely almost entirely on the US-provided technology stack. Consider Brazil. Its ICT market is $141.7 billion, 6.5 per cent of GDP (per Brasscom, Brazil’s ICT industry association), yet only 24.8 per cent of its software is domestically produced, per the BRICS Digital Sovereignty Index Report (citing ABES 2024). The cloud computing market is divided among Amazon, Microsoft, and Huawei. When data is stored on foreign companies’ servers under foreign legal jurisdiction, ‘sovereignty’ over that data amounts to a paper claim. South Africa’s situation is even more alarming. Scoring low on digital sovereignty assessments, South Africa is not pivoting toward autonomous construction; its EEIP framework, presented as a general policy for multinational ICT firms, would also address one of the principal regulatory obstacles to Starlink’s entry into the South African market.

2. Digital governance without sovereignty

Many Global South countries have enacted digital governance legislation, yet these measures have done little to alter the underlying relations of technological dependence. Brazil passed its General Data Protection Law (LGPD) in 2020; India enacted its Digital Personal Data Protection Act in 2023. But the gap between legal text and practical effect is enormous. Brazil’s detailed rules on international transfers of personal data only arrived in August 2024, with the ANPD’s Resolution CD/ANPD No. 19/2024 — years after the rest of the law took effect. India’s data protection law allows data to flow freely to any country except those on a government ‘blacklist’. Given US tech companies’ total penetration of India’s digital economy, this is an open door.

A 2020 World Economic Forum white paper went so far as to argue that governments only need ‘remote access’ to data held by companies; where the data is stored does not matter. The substance of this proposal is to maintain the status quo, ensuring Global South countries continue to hand all their data to US tech giants. Passivity in governance rules stems from asymmetry of power. When your infrastructure depends on others, your authority to make rules is limited.

3. Systematic erosion of domestic digital capabilities

This may be the most fundamental challenge. Brazil once pursued an ambitious strategy to build a domestic computing industry. Its 1984 Informatics Law reserved much of the domestic market for Brazilian firms, covering computer hardware, software, databases, and other digital products. By the late 1980s, this strategy had helped create a sizeable domestic computer sector. But under the Collor government in the early 1990s, trade liberalisation and the dismantling of market-reserve policies exposed local firms to foreign competition before they had reached technological maturity. Brazil was thus integrated into the global digital economy increasingly as a market for imported technologies rather than as a producer of them.

India followed a different path but reached a similar outcome. Semiconductor Complex Limited, approved in 1976 and producing chips by 1984, represented an early attempt to build indigenous semiconductor capacity. A major fire in 1989 destroyed much of the Mohali facility and set back the project, but the deeper problem was the absence of sustained state investment and industrial strategy after that rupture. As Taiwan, South Korea, and later China invested heavily in semiconductor fabrication, India’s economic reforms increasingly prioritised software and IT services over manufacturing. The consequences remain visible today: while China’s leading foundries are producing chips at 3-nanometre and below, India’s domestic fabrication capability remains concentrated in legacy process nodes, while advanced chips continue to be manufactured abroad. India became a global centre for software labour while remaining dependent on foreign firms for advanced semiconductor manufacturing.

In both cases, integration into global value chains occurred through subordinate positions: Brazil as a market for foreign digital products, India as a supplier of software services without control over the hardware base. The result was not simply technological backwardness, but the erosion of complete national digital ecosystems.

Industrial hollowing-out leads to brain drain, brain drain leads to declining policy judgment, declining judgment allows Western consulting firms to easily dominate the policy agenda. But what is most worrying is the shift in industrial elites’ own thinking. Nandan Nilekani, Chairperson of Infosys (one of India’s largest IT services firms), stated publicly at Meta’s ‘Build with AI’ summit in Bengaluru in 2024: ‘Our goal should not be to build one more LLM [large language model]. Let the big boys in Silicon Valley do it, spending billions of dollars. We will use it…’ The CEO of Tata Consultancy Services (TCS), India’s largest IT services company, said something similar. When the leaders of a country’s largest IT companies consider fundamental R&D to be someone else’s business, digital sovereignty is out of the question.

The three challenges reinforce each other. Infrastructure dependence weakens the material basis for autonomous governance. Passive acceptance of governance rules compresses the space for industrial policy. Digital capability gaps fundamentally undermine the capacity to recognise and address the first two problems. Together they constitute a structural trap.

You Cannot Change What You Cannot Measure

Digital sovereignty varies enormously among Global South countries, and a blanket ‘South versus North’ narrative cannot capture this variation. To formulate effective policy, a measurement tool is needed, one that can systematically diagnose each country’s digital sovereignty condition.

The Digital Sovereignty Index (DSI) is such a tool. Developed in 2025 by Xiong Jie — senior researcher at Tricontinental and Secretary-General of the Global South Academic Forum — the DSI decomposes digital sovereignty into four dimensions and sixteen specific indicators, each evaluated on a five-level maturity scale.

Data ownership autonomy is the concentrated expression of digital sovereignty. Without ownership of data, nothing else is possible. This dimension includes four indicators: data ownership legislation (whether the state has established a clear legal framework for data property rights), domestic data storage requirements (whether critical data must be stored within national borders), cross-border data flow protection (whether adequate safeguards exist when data leaves the country), and data value public benefit inclusion (whether value generated by user data flows to the public rather than being extracted as private profit by foreign platforms).

But data ownership cannot be realised in a vacuum. It requires the support of digital infrastructure autonomy. This dimension examines the degree of independence across four layers: basic hardware (chips, servers, storage devices), basic software (operating systems, databases, middleware, cloud platforms), application software, and information security.

Digital space governance autonomy ensures that a country can shape, not merely accept, the rules of digital space. Those rules are currently written through a set of bodies where the US holds structural influence: the Internet Corporation for Assigned Names and Numbers (ICANN) over domain names and internet addressing, the Internet Engineering Task Force (IETF) over core protocols, the Institute of Electrical and Electronics Engineers (IEEE) over technical standards, and the World Trade Organization’s Agreement on Trade-Related Intellectual Property Rights (TRIPS) over intellectual property — the same framework Prashad names as the ‘New International Property Order’. This dimension measures a country’s capacity to legislate domestically and to contest those international arenas rather than inherit their outcomes.

All three dimensions above depend on digital capability autonomy. This dimension assesses cutting-edge technology R&D, university STEM talent cultivation, industrial engineering capacity, and the degree of coordination between digital technology and national development strategy.

A clear logical relationship exists among the four dimensions. Data ownership autonomy is the concentrated manifestation of digital sovereignty, but its realisation requires infrastructure autonomy as a material foundation and governance autonomy as an institutional guarantee. All three depend on digital capability autonomy as the fundamental support. The structure of the DSI framework itself reveals the operating mechanism of the structural trap: if digital capability is insufficient, infrastructure and governance cannot be autonomous; if infrastructure is not autonomous, data ownership can only be a paper claim.

Each indicator uses a five-level maturity scale: Level 1 ‘Initial’ (the issue of autonomy in this area has not been recognised); Level 2 ‘Aware’ (the importance has been recognised, initial actions are being taken); Level 3 ‘Developing’ (active progress is underway, but significant dependence remains); Level 4 ‘Competent’ (strong international competitiveness); Level 5 ‘Independent’ (largely autonomous, with little constraint from other countries).

In March 2026, the International Communication Research Institute at East China Normal University, the Global South Academic Forum, and the Institute for Digital Economy & Artificial Systems (IDEAS) officially released the BRICS Digital Sovereignty Index Report at the Zhongguancun Forum. The findings are striking: China leads across all four DSI dimensions; Russia and India show strength in specific areas. But most of the newly admitted BRICS member countries — drawn from the broader Global South — remain at the earliest stages on infrastructure and core technologies. The structural trap described above is not an abstraction. It is what the numbers show.

The Same Label, Different Realities

The BRICS Digital Sovereignty Index Report’s assessment results reveal a critical fact: beneath the uniform label of ‘Global South’, countries’ digital sovereignty conditions differ dramatically.

China (DSI average 4.25) is the only country besides the United States with relatively complete digital sovereignty. Most indicators reach ‘Competent’ or ‘Independent’ levels. China’s relative weakness lies in international digital space rule-making, though its influence in international standards organisations has been growing steadily.

Russia (3.25) presents a distinctive case. Western geopolitical pressure and sanctions have pushed Russia to pursue digital sovereignty more aggressively. Russia scores highly on application software (5/Independent), basic software (4/Competent), information security (4), and talent cultivation (4). But it faces a severe bottleneck: chips. Basic hardware autonomy scores only 3 (Developing), with heavy reliance on imports. In 2021, Russia held just 1,973 international patents, 0.35 per cent of the global total. This figure dropped sharply in 2022–23 under Western restrictions.

India (2.94) is a classic case of one strong leg. The digital capability dimension is its bright spot: 2.55 million STEM graduates in 2020, second globally behind China, with a massive IT services ecosystem. But India is severely lopsided: focused on the application layer, doing almost no fundamental R&D. Core hardware and basic software depend heavily on US supply. Data protection legislation was only recently enacted and remains untested. Government investment is limited; brain drain is severe.

Brazil (2.13) appears fragile across all dimensions. After abandoning its domestic ICT industry, both the industrial base and talent pool are thin. Building independent digital infrastructure in the short term would be extremely difficult. The DSI report states bluntly: Brazil’s digital sovereignty remains fragile, highly dependent on Western companies.

South Africa (1.94) has relatively complete data protection legislation and a clear digital strategy on paper, but the gap between paper and reality is vast: weak enforcement, deep dependence on foreign core infrastructure, and domestic R&D constrained by limited resources and brain drain. Policy intent and legal frameworks fall far short of achieving digital sovereignty. Rather than pivoting toward autonomous construction, South Africa’s policy framework would ease Starlink’s entry into its market — suggesting that the political will to pursue digital sovereignty is itself weak.

Placed side by side, several judgments emerge. China is the exception; the rule is that most Global South countries’ digital sovereignty falls far below what outsiders might imagine. Patent data is especially stark: the US holds 21.11 per cent of global patents, China 39.84 per cent, Russia 0.35 per cent, Brazil 0.04 per cent, South Africa 0.01 per cent. The STEM graduate gap is equally telling: China 3.57 million, India 2.55 million, US 820,000, Russia 520,000, Brazil 238,000. The DSI assessment quantifies a reality that has long been treated with vagueness, forcing into view what the dominant powers have preferred to leave unmeasured: the full depth of the Global South’s digital dispossession.

The Window Is Closing

The DSI assessment does more than diagnose the present; it points toward a grim prospect. For most Global South countries, the historical window for independently building a complete ICT industry is narrowing. The capital threshold for the ICT industry is extreme. China and the United States each invest hundreds of billions of dollars annually in AI R&D. India’s national AI programme (IndiaAI) has a budget of $1.25 billion. The gap is two orders of magnitude.

But a narrowing window does not mean no options exist. Sinha notes that China’s development strategy centres on production, not rent extraction. Through long planning horizons, state coordination, mass technical education, and disciplined capital allocation, China has systematically built complete industrial ecosystems across multiple advanced sectors simultaneously. The socialist state has prevented domestic capital from consolidating into monopoly forms capable of extracting sustained super-profits. Firms are compelled to compete on cost, quality, and process innovation rather than relying on intellectual property rents. As a result, China has repeatedly transformed technologies that the imperial core treated as rent-generating monopolies into competitive, low-cost commodities.

This means Global South countries face a choice between two different logics. The US system operates on technological monopoly and rent extraction, using intellectual property regimes, trade agreements, and ‘multi-stakeholder’ governance frameworks to lock the Global South into permanent payment and permanent dependence. The alternative — demonstrated by China’s own development path — is production diffusion and technological democratisation: compressing monopoly rents into competitive costs, transferring capabilities rather than licensing access to them, building industrial ecosystems through state coordination rather than market extraction. South-South cooperation organised around this logic is not another form of dependency; it is the only credible path through a closing window.

The value of the DSI lies in enabling Global South countries to see their own full picture: which dimensions have foundations, which are weak points, where cooperation space exists, which links must remain under autonomous control. Measurement is the precondition for action. With diagnosis comes the possibility of strategy.

In the past, producing a podcast meant choosing a topic, writing a script, inviting guests, recording audio, and spending hours on post-production editing.

For independent creators this was extremely time-consuming. For small teams with limited capacity — no dedicated person to record and produce episodes — podcasting often fell off the agenda altogether. As a result, many valuable ideas were never turned into podcast episodes.

AI changes that.

What Is an AI Podcast?

Simply reading an article aloud is nothing more than text-to-speech.

An AI podcast goes much further. You feed in source material — PDFs, webpages, interview transcripts, and other documents — and the AI first understands the content, identifies the key ideas, reorganises them into a natural conversational script, and finally converts that script into an engaging audio programme.

The most valuable part of the process is therefore not the synthesised voice itself, but the content organisation that happens beforehand. A well-made AI podcast sounds like two people discussing ideas they have already digested and understood, rather than a machine reading directly from a document.

Why Create AI Podcasts?

Research papers, technical blogs, course notes, and industry reports often stay locked in text, out of reach during a commute, a workout, or any other moment suited to audio. When the material piles up, reading all of it is hard to keep up with. Turning it into podcasts lowers the barrier to consuming information — you can listen instead of read.

AI podcasts also make content reuse much easier. A single document or research report can become notes, a blog article, and eventually a podcast, without rewriting everything from scratch each time.

And they dramatically reduce production costs, handling most of the repetitive work so you spend far less time recording, re-recording, and editing audio.

None of this is meant to replace traditional podcasts — it simply puts podcasting within reach when producing one the traditional way isn’t an option.

AI Podcast Platforms

Today there are already several mature AI podcast solutions available. For most use cases, third-party tools with well-designed workflows can already produce podcast episodes quickly and consistently.

NotebookLM

The best known and most capable of these is NotebookLM.

Its Audio Overview feature can automatically generate a podcast-style discussion from PDFs, webpages, video transcripts, and other uploaded materials.

You can access this feature through the NotebookLM web application.

Open the NotebookLM website at <https://notebooklm.google.com/> and click Create new notebook.

Paste a webpage URL into the dialogue box at the top, or upload downloaded documents into the NotebookLM workspace. In this example, we use Chapter 1 (pages 27–59) of Karl Marx’s Capital: A Critique of Political Economy, Volume I, published in 1867.

After the file is uploaded, NotebookLM automatically generates a summary, providing a quick overview of the document before you begin exploring it.

Hover over the Audio Overview button to see that NotebookLM can automatically transform the uploaded material into a podcast using AI.

Depending on the size of the source material, generation may take several minutes. For this 33 page excerpt, the process takes approximately 10 minutes.

Once generation is complete, the audio file appears in the lower-right corner. Click the play button to listen to the podcast.

NotebookLM also offers an Interactive Mode, making the listening experience much more engaging. Instead of passively listening, you can interrupt the AI hosts and ask questions at any point during the discussion.

Click Join to enter the conversation. The AI hosts pause the discussion, answer your question, and then seamlessly continue where they left off.

The podcast generated by NotebookLM carries a natural rhythm and clear delivery, weaving a paper’s core ideas into a fluid two-person dialogue. Compared to reading a lengthy research paper from start to finish, this format is far better suited to quickly absorbing the key takeaways during spare moments.

You can listen to the full podcast created by NotebookLM in the Bandung Circuits repository.

Open Source AI Podcast Generators

Compared to mature AI podcast platforms like NotebookLM, the greatest advantage of open-source projects lies in their higher degree of freedom and stronger controllability. Users can freely choose from different large language models according to their own needs, striking a better balance among generation quality, response speed, and cost.

At the same time, most open-source projects support local deployment, so data does not need to be uploaded to third-party platforms, making them more suitable for scenarios that require privacy, security, or corporate data compliance. In addition, such projects usually support custom podcast personas, automation workflows through APIs, and allow developers to extend features or modify source code based on actual needs, unrestricted by commercial platforms.

If you want to delve deeper into AI podcasts, carry out secondary development, or build your own AI podcasting system, open-source solutions undoubtedly offer greater playability and room for expansion.

ai-podcast-creation Skill

The simplest open-source option is the ai-podcast-creation skill. Unlike NotebookLM, which produces a finished audio podcast, it generates the podcast script only. You can then bring that script to life however you prefer — record it yourselves, or convert it to audio with a text-to-speech tool. This suits a team whose hosts still want to voice the episode but would rather save the time writing the script.

Installing the skill is straightforward. Open Agent and enter the following command:

Please help me install this skill:
npx skills add inference-sh/skills@ai-podcast-creation

After installation finishes, restart VS Code to load the skill into your session. You can then generate a podcast script using a prompt such as:

/ai-podcast-creation Please generate a podcast for @project/AI-podcast/Capital-Volume-I-pages-27-59.pdf

The primary purpose of this skill is to generate podcast scripts, with the final output presented as a natural conversation between multiple hosts.

The full script is available in the Bandung Circuits repository.

Open Notebook

Open Notebook has a very clear positioning: it aims to become the open-source alternative to NotebookLM.

Officially, it is defined as an AI platform for research, learning, and knowledge management, not merely a podcast generation tool. As a result, besides supporting AI podcast generation, it also offers common NotebookLM features such as AI Q&A, AI notes, and AI summarisation.

More importantly, Open Notebook is also very friendly to ordinary users. The project provides a complete Web UI, so even without any programming background, you can generate AI podcasts through a visual interface, just like using ordinary desktop software.

Before installing it, there is one prerequisite worth understanding: because Open Notebook runs on your own computer but still relies on external AI models, you will need an API key.

What Is an API Key?

If you plan to try out local AI projects, an API key is essentially unavoidable. Although many AI applications run locally, the core capabilities that actually consume computing resources and determine output quality usually still come from third-party AI services.

Take AI podcasting as an example: the entire generation pipeline typically involves multiple steps such as article comprehension, content summarisation, script generation, and speech synthesis. If you rely entirely on local models to complete these tasks, not only will you need a fairly high hardware configuration, but the generation speed and final quality will often fall short compared to cloud-based models, making the overall cost-effectiveness quite low.

Therefore, most projects will call model services such as OpenAI, Google, Anthropic to handle these core tasks, and an API key is essentially the key to accessing these services. Only after configuring an API key in your local project can you successfully invoke the corresponding models. Of course, model providers will charge API fees based on actual usage.

Installing and Configuring Open Notebook

It is recommended to use an AI agent to automatically complete the installation and configuration of Open Notebook. After opening VS Code, simply enter the prompt below.

Set up Open Notebook on this computer.

If Git is installed, clone https://github.com/lfnovo/open-notebook.git. Otherwise, download https://github.com/lfnovo/open-notebook/archive/refs/tags/v1.10.0.zip and extract it.
Follow the instructions in README.md to install dependencies and start the application.

After the installation completes, the Web UI will usually open automatically. If it does not start automatically, you can visit http://localhost:8502 in your browser.

Click the Models button in the lower left corner to enter the model configuration page.

Open Notebook supports multiple model service providers. Below we will use OpenAI as an example for configuration.

First, go to the OpenAI platform to create an API Key.

https://platform.openai.com/api-keys

Note that the OpenAI API and ChatGPT Plus subscription are not interchangeable; the API service requires a preloaded balance before it can be used normally.

https://platform.openai.com/settings/organization/billing/overview

Click Create new secret key to create a new API Key.

Fill in the API Key name, then click Create secret key.

The API Key will be shown only once, so it is recommended to copy and save it immediately. Its format is usually sk-xxxxx.

Then return to Open Notebook and click Add Configuration.

Fill in the Configuration Name and paste the API Key you just saved to complete the setup.

You can click Test to check whether the API can be called successfully. If the test fails, first verify that your API account balance is sufficient.

After confirming the configuration is successful, click Models to start adding models.

Open Notebook requires you to configure four types of models:

• Language (Large Language Model): Responsible for text tasks such as conversation, summarisation, reasoning, and content generation.
• Embedding (Text Embedding Model): Responsible for converting text into vector embeddings, used for RAG, semantic search, similarity computation, and related functions.
• TTS (Text-to-Speech): Responsible for converting text into speech.
• STT (Speech-to-Text): Responsible for speech recognition and transcription.

Below that, all models currently supported by OpenAI will be displayed.

If you are unfamiliar with the differences between these models, you can hand the model list over to AI and let it help recommend the most suitable configuration. (Replace [Model List] below with the actual model list.)

I need to select the most appropriate model for each of the following four task categories from the model list below, and I would like you to explain your recommendations.

The task categories are:

**Language (LLM):** General-purpose text generation, including conversational AI, question answering, reasoning, summarization, and content generation.

**Embedding:** Converting text into vector embeddings for semantic search, retrieval-augmented generation (RAG), similarity search, clustering, and related applications.

**TTS (Text-to-Speech):** Converting text into natural-sounding speech.

**STT (Speech-to-Text):** Transcribing spoken audio into text.

For each task category, please:

- Recommend the best-suited model.
- Briefly explain why it is recommended, considering factors such as capability, performance, latency, and cost.
- Clearly indicate if no suitable model exists for a particular category.
- If any listed model is deprecated or outdated (for example, `text-embedding-ada-002`), recommend the most appropriate modern replacement instead.

Using the following model list:
[Model List]

In this article, we will use the model combination recommended by GPT to generate a podcast:

• Language: gpt-5.5
• Embedding: text-embedding-3-large
• TTS: gpt-audio
• STT: gpt-4o-transcribe

For example, set Model Type to Language, choose gpt-5.5, then click Add to add the model.

After adding, you can select gpt-5.5 from the Chat Model dropdown above.

Follow the same method to add all the models required by Open Notebook one by one.

Then go to Podcasts → Profiles to configure the corresponding models for your podcast presets.

Here we take tech_discussion as an example; click Edit.

For all configuration items marked with *, select the models you have already added, such as gpt-5.5.

After completing the configuration, check whether the right side still shows Setup required. If this reminder disappears, it means the current template has been fully configured.

Continue configuring the remaining templates until all Setup required indicators have disappeared; then you can officially start generating podcasts.

Creating a Notebook and Importing Articles

First, click New → Notebook to create a new notebook.

Enter the notebook name and click Create New Notebook.

After creation, you will see the new notebook on the Notebooks page on the left.

Enter the notebook and click Add Source to import the article you want to convert into a podcast.

Open Notebook supports three import methods: entering a web page URL, uploading a local file, and pasting text directly. This article demonstrates the second method by uploading a local file.

After the system finishes parsing, the imported article will appear in the Sources section, indicating that the material has been successfully indexed.

Generating an AI Podcast

After importing the materials, click New → Podcast again to create a new podcast project.

The Episode Settings on the right allows you to set the podcast style. Open Notebook provides three default presets:

• business_analysis: A business analysis, financial interpretation, and case study style, generally more formal.
• solo_expert: A solo expert presentation, similar to a lecture or a teacher’s lesson.
• tech_discussion: A two-host discussion on technology, papers, AI, programming, and similar topics, closer to NotebookLM’s podcast style.

In addition to the presets, you can add extra instructions in Additional instructions, such as requesting more humour, adjusting speech rate, controlling the episode length, or specifying the style of expression.

After confirming the settings, click Generate, and Open Notebook will start generating the podcast.

Checking Generation Progress

After the generation task is submitted, click the Podcasts page on the left to view the current task list.

Depending on the article length and model response speed, the entire generation process usually takes a few minutes. When the task is completed, the corresponding podcast will appear in the list.

After the podcast conversion is complete, click the play button at the bottom of the page to listen to the AI podcast generated by Open Notebook online.

If you are not satisfied with the podcast content, host style, or delivery, you can return to Episode Settings, adjust the preset template or modify the Additional instructions, and then regenerate.

Since Open Notebook supports freely choosing large language models and speech models, swapping different model combinations can also yield vastly different podcast results. For users who want to build a personalised AI podcast workflow, this high degree of configurability is one of the greatest advantages of the open-source approach.

You can listen to the full podcast created by Open Notebook in the Bandung Circuits repository.

Conclusion

AI has fundamentally changed the podcast production workflow. Instead of spending most of your time on scripting, recording, and editing, you can now begin with existing documents, allow AI to understand and reorganise the material, generate a conversational script, and finally produce a polished audio programme.

For most creators, the greatest value is not the synthesised voice itself, but AI’s ability to transform complex written material into engaging conversations. Whether you are working with research papers, technical documentation, course materials, or industry reports, AI makes it possible to repurpose knowledge into an accessible audio format with far less effort than traditional podcast production.

Assessment prepared in accordance with the methodological specifications of the Digital Sovereignty Index (DSI v2.0), under the academic cooperation framework of the Global South Academic Forum.

Chapter 1 — Executive Summary

Argentina exhibits a paradoxical pattern of digital sovereignty: a regulatory scaffolding and a software-engineering capability that are comparatively sophisticated for the region coexist with a deep structural dependence on infrastructure, institutional enforcement capacity and capability autonomy — all of it cross-cut by a trajectory of active deregulation that operates as a ceiling factor during the assessment period (2024–2026, Milei administration). The country knows how to legislate on data, and to export software services, but it does not capture the economic value of its data, it does not enforce with deterrent effect, and, in the cross-border flow, it has ceded sovereignty through the adequacy commitment with the United States. The result is a mid-to-low-grade digital sovereignty, without a single indicator reaching the competence level (Level 4), and with two indicators collapsed to the initial level (Level 1).

On the DSI framework’s 1–5 scale, Argentina obtains a sovereignty score of 2.00 / 5.0 (a total of 32 out of 80). The score is computed as the arithmetic mean of the sixteen indicator ratings; no indicator exceeds Level 3. The distribution by dimension reveals that the weakness is concentrated in data ownership.

Table 1.1. Indicator ratings (UNGS_2026 corpus).

Table 1.2. Averages by dimension.

Three principal strengths. First, a mature and long-standing data-protection framework: Law No. 25326 (Personal Data Protection Act), enacted in 2000, has been in continuous force, operated by a functional enforcement authority, and Argentina has retained the European Union’s adequacy status since 2003 ¹ — indicator 1.1 is the only one in the entire matrix that reaches Level 3 on its own legislative merit. Second, a software industry of international scale: revenue of USD 22,221 million and more than 158,000 registered jobs in 2024, with a domestic fintech subsector that commands 88% of digital banking ² — indicator 4.3 is the second and last Level 3. Third, an installed base of policy recognition and bounded technical capacity in Dimension 2: a state GNU/Linux distribution, a national public cloud on open-source code, and citizen platforms authored by the State ³.

Three principal weaknesses. First, the collapse of cross-border flows: the commitment to recognise the United States as an adequate jurisdiction, formalised in November 2025 and February 2026, brings indicator 1.3 down to Level 1 through the mechanism of Principle #7 ⁴. Second, the total absence of data value capture: Argentina gives its public data away for transparency but possesses no fiscal instrument, no B2G mandate, and no recognition of data as an economic asset, which places indicator 1.4 at Level 1 ⁵. Third, a cross-cutting structural dependence in Dimensions 2 and 4: the country does not manufacture hardware, does not control its system-software stack, and exports its most capable talent abroad ⁶.

The cross-cutting finding that orders the entire assessment is the Milei administration’s deregulation trajectory as a ceiling factor. This is not a context that explains low ratings, but an active policy that subtracts capacity: 543 measures that modify or eliminate 2,519 norms across 15,144 articles ⁷, a cumulative real cut of 50.6% in the science-and-technology budget function ⁸, and the cession of sovereignty over data flows. Where other countries build, Argentina dismantles — and the DSI framework, through its trajectory clause and its Principle #7, registers that direction as a ceiling, not as a mitigating circumstance.

Chapter 2 — Methodological Framework

2.1 The DSI 4×4 framework

The Digital Sovereignty Index evaluates the degree to which a country has attained independence in the digital domain: the extent to which its data, its infrastructure, its governance and its capabilities are determined by the country itself and not by foreign actors, jurisdictions, or platforms. The operational definition, set out in the DSI methodological specifications, rests on the word ‘independence’: the index does not measure whether a country aspires to be sovereign, but whether, in fact, it is. The framework organises the assessment into a matrix of four dimensions by four indicators — sixteen indicators in total — defined in the Digital Sovereignty Index specifications coordinated by Global South Insights through the DSI Assessment Team, under the Global South Academic Forum. This Argentina assessment was prepared by UNGS as an associate university, applying those methodological specifications to the national situation: evidence collection from Argentine sources, the local implementation of the assessment pipeline, and the drafting of the analysis in academic Spanish constitute UNGS’s contribution to the global index compiled by GSI.

The four dimensions are: Dimension 1 — Independence in data ownership (1.1 Data-ownership legislation; 1.2 Domestic storage; 1.3 Protection of cross-border flows; 1.4 Data value for public benefit); Dimension 2 — Digital infrastructure independence (2.1 Basic hardware; 2.2 System software; 2.3 Application software; 2.4 Information security); Dimension 3 — Digital governance independence (3.1 Legislative capacity; 3.2 Enforcement capacity; 3.3 Leadership in international technical rules; 3.4 Leadership in international conduct rules); and Dimension 4 — Digital capacity independence (4.1 Frontier research; 4.2 University talent; 4.3 Industrial engineering; 4.4 National strategic alignment).

2.2 The 1–5 scale

Each indicator is rated on a scale of five progressive levels of independence: 1 Initial (the issue has not been addressed; potentially complete dependence), 2 Aware (the importance of independence is recognised and discussions or actions have been initiated, typically at the planning stage), 3 Developing (work is actively underway towards independence, with policies implemented but still with strong external dependence), 4 Competent (international competitiveness and full potential autonomy) and 5 Independent (basic self-sufficiency, with minimal external constraints). The scale is applied per indicator and prohibits decimal ratings: each level must be tied to a defensible path through a four-node decision tree (existence of critical evidence → effective implementation → meaningful enforcement → international competitiveness vs. self-sufficiency).

2.3 Principles, pitfalls, and the decision tree

Rating is governed by seven immutable principles that prevail over the decision tree when they come into tension with it: Facts over Law (effective implementation prevails over legal text), Effectiveness Supreme (laws that cannot be enforced receive lower ratings), Pragmatism (real status, not declared goals), Dependency Penalty (dependence on foreign platforms caps the ceiling), Corporate-Capture Degradation (legislative blocking by Big Tech lowers the rating), Context Explains but Does Not Excuse (geopolitical pressure explains but does not modify consequences), and the Digital Hegemony Reality Check — Principle #7 (designating the United States as an ‘adequate jurisdiction’ significantly lowers ratings). To the principles are added eight pitfalls, or rating traps — among them confusing activity with outcomes, underestimating weaknesses, and geographic presence without control — that must be cleared before finalising a rating, together with a horizontal-consistency check against the anchors of China, Russia, and Brazil.

2.4 The UNGS 2026 corpus

The assessment is grounded in the UNGS_2026 corpus, built by the UNGS agent pipeline. The collector gathered 373 raw pieces of evidence across the three search rounds; the verifier removed 91 for verification failures (invalid URL, lack of content match, or suspected hallucination); the integrator consolidated the remainder by removing 41 duplicates through exact URL match, leaving 241 integrated pieces of evidence (AR-EV-001 to AR-EV-241) plus 31 documented gaps. The composition by type covers regulation, case studies, quantitative data, reports, analysis, policy, and gaps; confidence is distributed across 125 high-confidence items, 80 medium, and 36 low. The share of Tier-1 sources (official domains *.gob.ar, InfoLEG, Boletín Oficial (Official Gazette)) reaches 48.5%, above the 40% minimum but below the 60% target. Each [AR-EV-NNN] citation in this report is traceable to a verified URL in evidence_base_UNGS_2026.json.

2.5 The Principle #7 override mechanism

One methodological element deserves anticipating because it proves decisive for indicator 1.3. Principle #7 does not function as a gradual downward adjustment but as a categorical-collapse mechanism: the active designation of the United States as an adequate jurisdiction for personal-data transfers relocates indicator 1.3’s rating to Level 1, overriding the result that the decision tree would have yielded on its own merits. The reasoning is that ceding adequacy to a jurisdiction without comprehensive federal privacy legislation and with statutory extraterritorial-access regimes (CLOUD Act) hollows out the protective purpose of the indicator, regardless of how sophisticated the rest of the framework may be.

2.6 Limitations

The assessment acknowledges four limitations. First, 31 candidate items were discarded as URL_BLOCKED owing to anti-bot barriers, TLS, or 403 responses (ITU documentary databases, 3GPP partner listings, W3C sources); human re-retrieval is recommended for future runs, since these are transport limits, not veracity limits. Second, indicators 3.3 and 3.4 fell below the sufficiency threshold of 15 pieces of evidence (11 and 12 respectively) after the verifier’s removals, which limits their confidence to Low. Third, several de facto quantitative metrics rest on single Tier-3 sources or on inference (hyperscaler market shares, the brain-drain rate specific to computer science), which reduces confidence without invalidating the direction of the rating. Fourth, two items suspected of hallucination were removed by the verifier and do not appear in the library; this report does not reproduce their claims, in keeping with the anti-hallucination discipline that governs the pipeline.

Chapter 3 — Indicator-by-Indicator Analysis

Dimension 1 — Independence in Data Ownership

Dimension 1 asks who owns, controls, and benefits from the data generated by Argentine residents, firms and institutions. It is the dimension where Argentina displays its sharpest internal contrast: it has the oldest and most recognised data-protection legislative framework in Latin America — the first regional adequacy status before the European Union, in 2003 ⁹ — but that historical asset coexists with a flow protection collapsed by US adequacy and with a categorical absence of value-capture mechanisms. The relevant institutional landscape includes the AAIP (Agency for Access to Public Information) as the enforcement authority and the Law No. 25326 regime as the central piece. The dimension finding is unequivocal: with an average of 1.75, data ownership is Argentina’s weakest dimension, dragged down by the simultaneous collapse of two of its four indicators to Level 1.

3.1.1 Indicator 1.1 — Data ownership legislation

Rating: 3 — Developing. Confidence: High.

Argentina has a general data-protection statute in continuous force since the year 2000 — Law No. 25326 ¹⁰ — operationalised by Decree 1558/2001, verified as in force, and amended on thirty-two occasions ¹¹, and administered by the AAIP, an autonomous statutory body with a permanent National Directorate for the Protection of Personal Data ¹². The authority is not nominal: it issues resolutions of operational substance — security measures, model contractual clauses for transfers, and the tiering of sanctions ¹³ — and exhibits a multi-year activity trace (491 case files and 52 sanctions in 2022) ¹⁴. The framework retains current external recognition: the European Union’s adequacy status was reaffirmed in January 2024 ¹⁵ and Convention 108+ was ratified by Law No. 27699 in 2022 ¹⁶. This conjunction of an operative statute, a functional regulator, and international recognition defines the Level 3 profile.

What prevents an ascent to Level 4 is the conjunction of two independent failures. First: the modernised text is not in force. The reform bill (Message 87/2023), which incorporated breach notification, impact assessments, portability, and the data-protection officer, lost parliamentary status at the end of 2024 after more than thirty months without consideration ¹⁷; the post-GDPR-era provisions remain absent from the text in force ¹⁸. By Principle #1 (Facts over Law), a bill cannot anchor a higher rating. Second: enforcement is not deterrent. The fine cap was never updated by law and stands at around USD 70–100; AAIP Resolution 126/2024 graduated infractions but could not raise the legal ceiling, because an administrative resolution does not amend a law ¹⁹; the total of 2022 fines was some USD 30,621 ²⁰ and the mass leaks of 2024–2025 (Renaper, ARCA, ANSES) were not sanctioned ²¹. By Principle #2 (Effectiveness Supreme), enforcement without deterrent magnitude is capped at Level 3. It does not descend to Level 2 because the statute is in force — it is not a draft — the regulator operates and there is an enforcement trace: Argentina comfortably clears the band of ‘preliminary discussions or planning’. In the comparative frame, 1.1 sits well below the Chinese anchor (Level 5) and groups with Brazil (Level 3): both regimes are transplanted from external standards and their enforcement scaling has lagged, although the Argentine weakness is specific — an obsolete text from 2000 whose modernisation stalled. An ascent to Level 4 would require the modernised text to enter into force and the fine cap to reach deterrent magnitudes on the order of 2–4% of revenue; erosion towards Level 2 would only follow the dismantling of the AAIP or the repeal of the regime in force.

3.1.2 Indicator 1.2 — Domestic data storage

Rating: 2 — Aware. Confidence: Medium.

Argentina lacks a general data-localisation mandate. The personal-data regime (Law No. 25326, art. 12) governs cross-border flows through adequacy and contractual clauses, not through territorial storage, and the AAIP’s own security resolution imposes no geographic restriction whatsoever ²². The only operative sectoral rule is the Central Bank’s prudential regime on third-party control and delegated technology (BCRA (Central Bank of the Argentine Republic) Communication A 7724, updated by A 8401) ²³, which structures the supervision of outsourced services but does not require storage in national territory. Two reform bills explicitly preserve the transfer-based model and decline to introduce a localisation mandate ²⁴. A residual sovereign capacity exists — ARSAT’s Tier III data centre and the national public cloud on open-source code ²⁵ — but it is niche, oriented to the public sector, and undergoing partial privatisation (49%) ²⁶. This configuration — a sectoral-only rule, residual domestic infrastructure, non-localising bills — corresponds precisely to Level 2.

The indicator does not reach Level 3 because that would require effectively enforced localisation in multiple substantive sectors plus a domestic cloud industry with measurable share: Argentina meets neither. The installed national capacity (~32 MW) is orders of magnitude smaller than a single hyperscale facility ²⁷, while the growth of the cloud market is captured by foreign providers — AWS’s Local Zone in Buenos Aires and the OpenAI/Sur Energy 500 MW project under the RIGI ²⁸ — a configuration that the framework rates as geographic presence under foreign jurisdiction (CLOUD Act), not as sovereignty (Pitfall #8). State policy actively incentivises foreign-controlled infrastructure without a localisation counterpart ²⁹. It does not descend to Level 1 because there coexist a sectoral mandate in force, an operative state alternative and documented legislative discussion — the presence of the three elements the criterion requires to clear the initial floor. Argentina groups with Brazil (Level 2): both lack a general mandate and see their markets dominated by hyperscalers, while Russia (Level 4, Law 242-FZ with documented enforcement) marks exactly the regime that the Argentine bills decline to adopt. An ascent to Level 3 would require a sectoral localisation mandate effectively audited in finance and health or government, together with a domestic cloud industry of non-residual share.

3.1.3 Indicator 1.3 — Protection of cross-border flows

Rating: 1 — Initial. Confidence: High.

On paper, Argentina possesses a comprehensive and operationally exercised cross-border transfer regime: article 12 of Law No. 25326 prohibits transfers to jurisdictions without adequate protection ³⁰, Decree 1558/2001 empowers the regulator to assess adequacy ³¹, Disposition 60-E/2016 publishes model contractual clauses ³², AAIP Resolution 34/2019 maintains a granular adequacy list from which the United States was explicitly absent ³³, and the country retains European adequacy status since 2003, revalidated in January 2024 ³⁴. This conjunction of an operative statute, a functional regulator and international recognition defines the Level 3 profile.

However, the Joint Statement of 13 November 2025 and the USTR Fact Sheet ³⁶, formalised by the signing of the Agreement on Reciprocal Trade and Investment on 5 February 2026 ³⁷, commit Argentina to recognising the United States as an adequate jurisdiction for personal-data transfers. By Principle #7 (Digital Hegemony Reality Check) and the indicator’s ‘collapse through extreme permissiveness’ mechanism, an active US adequacy designation caps the rating at Level 1, overriding the narrowing of the decision tree. Sophistication on paper does not anchor the rating; the operational reality of structural deference to a hegemonic jurisdiction does (Principle #1). This is a Level 1 by way of collapse, not by regulatory void: the regime exists and is exercised, but the adequacy designation to the dominant jurisdiction of data platforms removes the only protective barrier — historically, US Big Tech operated via contractual clauses, not by adequacy ³⁸. In the comparative frame, the decisive separator with respect to China (≈5) and Russia (≈4) is exactly the variable that Principle #7 isolates: neither anchor grants adequacy to the United States, while Argentina now does. The result coincides with the BRICS 2025 baseline (also Level 1) by identical collapse reasoning. The only possible upward move would be the non-entry-into-force or the reversal of the US adequacy commitment; without it, no level above 1 is attainable.

3.1.4 Indicator 1.4 — Data value for public benefit

Rating: 1 — Initial. Confidence: High.

Argentina sustains a mature and long-standing open-data programme — Decree 117/2016, Law No. 27275 on access to information, and the datos.gob.ar portal with 1,235 datasets from 42 agencies ³⁹ — together with an active civic-tech ecosystem that reuses public data ⁴⁰. But that programme operates entirely on a logic of transparency and reuse, not of economic value capture from the data. The conceptual distinction is decisive: a robust open-data programme ‘gives data away to enable innovation’, while the indicator measures ‘capturing the value of data for public benefit’. All of Argentina’s substantive positive evidence belongs to the first category.

The indicator’s critical enablers are absent and documented as gaps with Tier-1 search: there is no state recognition of data as a factor of production or economic asset ⁴¹, there is no operative B2G mandate requiring platforms to share datasets of public interest ⁴², there is no fiscal value-capture instrument — the PAIS tax was a consumption tax, excluded by criterion, and was moreover repealed in December 2024 ⁴³, and the OECD’s Pillar One and Pillar Two were not implemented ⁴⁴ — antitrust is not applied to data monopolies — the CNDC (National Commission for the Defence of Competition) maintains a study group without sanctioning power, and none of the eleven concentrated markets under investigation is a digital-native platform ⁴⁵ — and there are no institutional data trusts ⁴⁶. Decree 780/2024, moreover, restricted the scope of active transparency: a regression, not an advance ⁴⁷. By Principle #3 (Pragmatism), aspirational bills do not anchor higher levels; the categorical absence of the critical piece ends the decision tree at Level 1. The indicator does not ascend to Level 2 because there does not even exist governmental recognition of data as an asset — not merely civil-society discourse — in a strategy or bill. The sophistication of the open-data programme places Argentina at the high end of Level 1, above a country with no policy at all, but transparency is not value capture. Against China (Level 4–5: data as the fifth factor of production since 2019), Argentina exhibits a lag of several levels; even against Russia and Brazil it remains a step below, lacking the limited B2G mechanisms and the digital-services taxation that those countries possess. An ascent to Level 2 would require the BCRA’s Open Finance System to enter into force or a bill to recognise data as an economic asset.

Intra-dimensional analysis. Dimension 1 reveals the Argentine paradox in its pure state: indicator 1.1, the strongest of the entire matrix on its own merit (the only framework with 2003 European adequacy), coexists with the collapse of 1.3 to Level 1 through the 2026 US adequacy, and with the total void of 1.4. The common bottleneck is that Argentina built capacity for protection without capacity for exploitation: it legislates the data point, but neither retains it territorially, nor protects it on its way out to the hegemonic jurisdiction, nor captures its economic value. The regulatory sophistication of 1.1 does not compensate for, and in fact contrasts with, the cession of sovereignty of 1.3 and the absence of economic vision of 1.4.

Dimension 2 — Digital Infrastructure Independence

Dimension 2 evaluates whether the country can operate its digital infrastructure without dependence on, or interruption by, foreign providers, across four layers: hardware, system software, application software and security. The relevant institutional landscape includes ARSAT and the national public cloud, the Tierra del Fuego regime, INTI, and INVAP on the technical plane, and CERT.ar in security. The dimension finding is notably uniform: all four indicators sit at Level 2 (average 2.00), a pattern of ‘awareness without substantive capacity’ in which Principle #4 (Dependency Penalty) — which the framework declares dominant for this dimension — operates across the board. Argentina recognises its dependencies and possesses real traces of capacity, but in each layer the structural dependence on the foreign stack caps the rating.

3.2.1 Indicator 2.1 — Basic hardware autonomy

Rating: 2 — Aware. Confidence: Medium.

There is explicit recognition of hardware dependence in national-level instruments — the Compre Argentino (Buy Argentine) regime (Law No. 27437) and the Knowledge Economy promotion regime (Law No. 27506) ⁴⁸ — and an at-scale capacity for assembling imported components under the Tierra del Fuego regime (Law No. 19640), where close to 93% of telephones, air conditioners, and televisions sold are assembled domestically ⁴⁹. But assembly is not manufacturing: by the principle ‘assembly is not autonomy’, assembling imported boards is value-added activity, not the substitution of foreign capacity. There exists no funded foundational hardware programme reaching the hinge threshold of ~USD 1,000 million over five years oriented to chip design or wafer fabrication: the RIGI attracts foreign investment in an enclave model without a technology-transfer obligation ⁵⁰, the Knowledge Economy benefits target software and services, not manufacturing ⁵¹, and the national science-and-technology strategy (CTI Guidelines 2025–2027) explicitly omits microelectronics and semiconductors ⁵². Semiconductor capacity is design-only: the INTI provides integrated-circuit design services that are synthesised in foreign foundries, and Argentina operates no wafer plant ⁵³.

The trajectory, moreover, is downward within the band: Decree 333/2025 reduces the import tariff on mobile phones from 8% to 0% by January 2026 ⁵⁴, Decree 111/2025 dilutes the Fuegian promotion fund, and the Mirgor Group suspended around 360 workers in response ⁵⁵. The indicator does not reach Level 3 because that requires a funded foundational programme in design or fabrication, or a domestic-supplier share of 10–30% in some hardware category designed in the country: Argentina meets neither. It does not descend to Level 1 because both policy recognition and an operative assembly capacity persist. In the comparative frame, Argentina sits alongside Brazil (Level 1–2, with its failed CEITEC plant) and on a par with the Russian anchor (Level 2); Level 3 in this region requires extraordinary evidence — a sustained fabrication programme with output measured in wafers. An ascent to Level 3 would require a funded foundational programme; erosion towards Level 1 would follow the complete collapse of the Fuegian assembly base.

3.2.2 Indicator 2.2 — System software autonomy

Rating: 2 — Aware. Confidence: High.

Argentina presents the canonical Level 2 pattern: policy recognition and demonstrated engineering capacity, without a funded national programme or measurable domestic share. There is explicit recognition of software sovereignty — the ONTI Public Software initiative, ONTI Disposition 2/2019 requiring code to be shared under open licences across the entire National Public Sector, and Santa Fe’s provincial law on the preferential use of free software ⁵⁶ — and a bounded implementation footprint: a state GNU/Linux distribution (Huayra, developed at EDUCAR) ⁵⁷, a national public cloud on open-source code ⁵⁸ and an ecosystem of official repositories (argob, 43 repositories) ⁵⁹. This satisfies the existence and implementation nodes, but no piece crosses the Level 3 thresholds.

Foreign dominance is overwhelming and quantified: the domestic desktop operating system is marginal (Windows 81.64%, total Linux 2.21%) ⁶⁰, the mobile market is more than 99.97% foreign ⁶¹ and the database market is dominated by Oracle, SAP, Microsoft, and IBM, with effectively zero domestic share ⁶². Decisively, ARSAT’s own state cloud runs its network operations centre on Red Hat OpenShift — control software from a US provider ⁶³ — so the control plane is not domestic: the physical location of the data centre does not equate to autonomy (Principle #7). The only domestic operating-system footprint (Huayra) was confined to the education sector and was suspended and intervened in the Milei era: the Conectar Igualdad and Educ.ar platforms were left in ‘construction mode’ and Decree 963/2024 appointed an intervenor for EDUC.AR S.E. ⁶⁴, downgrading that single case to symbolic status. Linux adoption is necessary but not sufficient, because its upstream is controlled from outside the country. It does not descend to Level 1 because a documented state distribution and a nationally scoped public-software regulation persist. Argentina is directly comparable to Brazil (Level 2) and remains a step below Russia (Level 3–4, with certified Astra Linux and sanctions-accelerated substitution). An ascent to Level 3 would require critical-infrastructure systems running on a domestic operating system with audited patching independence, or more than 5% domestic share in cloud, databases, or server operating systems.

3.2.3 Indicator 2.3 — Application software autonomy

Rating: 2 — Aware. Confidence: Medium.

Argentina presents the textbook Level 2 profile for this indicator: a capable software industry and genuine domestic dominance in several vertical categories, against complete foreign control of every horizontal category. On the vertical plane the evidence is strong and verified: Mercado Libre led national app downloads in 2024 (11.7 million) ⁶⁵, Mercado Pago surpassed 50 million monthly active users in Latin America with USD 8,600 million in revenue ⁶⁶, three domestic neobanks retain 88% of digital banking ⁶⁷ and the State deploys self-authored citizen platforms at scale — Mi Argentina, with 21 million registered users ⁶⁸. On the horizontal plane, dependence is almost total and quantified: WhatsApp with 93% penetration in messaging, Instagram with 86.7% ⁶⁹, Google with 94.01% of searches ⁷⁰, and productivity, ERP, and CRM dominated by Microsoft, SAP, Oracle, and Salesforce with no domestic provider of measurable share ⁷¹.

The decisive test of the indicator is horizontal share: an ascent to Level 3 requires ≥30% domestic share in at least one major horizontal category (office software, ERP, or industrial software). The Argentine shares above 30% are all in vertical categories — e-commerce, payments, digital banking — which the criterion explicitly excludes from the horizontal denominator. Policy is one of promotion, not substitution: Law No. 27506 sustains a service-export industry (USD 2,674 million in 2024, 46.9% to the United States) ⁷² rather than the substitution of horizontal products, and a domestic champion in a horizontal-adjacent category (Auth0, authentication) was acquired by the US firm Okta, reversing the sovereignty relationship (Principle #5) ⁷³. It does not descend to Level 1 because there are measurable domestic products, policy recognition, and a substantive software industry. Argentina groups with the lower band of Brazil but below it, because Brazil has, in TOTVS, a domestic ERP provider with measurable horizontal share — exactly what Argentina lacks — and remains a level below Russia (Level 3, with 1C in ERP and MyOffice in office software). An ascent to Level 3 would require a domestic provider to reach 30% share in a horizontal category or the public-procurement preference for nationally originated software to be applied with auditing.

3.2.4 Indicator 2.4 — Information security autonomy

Rating: 2 — Aware. Confidence: Low.

Argentina possesses a comparatively rich cybersecurity institutional architecture but fails the three substantive tests that separate Level 2 from Level 3. There is a national CERT in regulation (CERT.ar, Disposition 1/2021) ⁷⁴, operative on paper — it registered 438 incidents in 2024, 15% above 2023, with 61% in the state sector ⁷⁵ — and internationally recognised through FIRST membership. The scope of that figure must be made explicit: CERT.ar’s mandate is confined to the National Public Sector, within a deliberately federated incident-response architecture. Alongside it operate a second national-level team at the Ministry of Security and a network of sub-national centres — among them the BA-CSIRT of the Autonomous City of Buenos Aires, the CSIRT of the Province of Buenos Aires, and that of the Province of Córdoba, several of them integrated into the OAS CSIRT Americas Network — each with its own jurisdiction and sectors. These teams do not aggregate into the national score of this indicator, which measures federal-level capacity; their existence, however, means that CERT.ar’s register must not be read as an exhaustive count of the country’s incident activity, but as the activity of the federal-level team. A layered regulatory regime governs the public sector: the PNICIC critical-infrastructure programme, minimum security requirements with a duty to notify the CERT within 48 hours (Administrative Decision 641/2021) ⁷⁶, a second National Cybersecurity Strategy and the reorganisation of Decree 941/2025 that creates the National Cybersecurity Centre, and the Federal Cyberintelligence Agency ⁷⁷. But, by Principles #1 and #2, strategies, and constitutive instruments are activity, not measured effectiveness.

The triple test fails simultaneously: the CERT’s capacity (438 total incidents — not ‘major’ — reported without standardised FIRST-type operational metrics — mean time to respond, severity classification, or a trace of advisories adopted; the count is a raw aggregate that, moreover, reflects only the National Public Sector perimeter, and does not evidence the handling of major incidents with disclosed timelines) ⁷⁸; the market test (no measurable domestic share and no documented sovereign cryptographic research) ⁷⁹; and the product test (no verifiable domestic security product in critical infrastructure) ⁸⁰. Three major breaches of the period are reverse evidence that the regime does not protect: the leak of the national driver’s-licence database (~5.7–6 million records, 1.25 TB, April 2024) ⁸¹, the December 2024 defacement of Mi Argentina and SUBE — where the attackers reported the absence of two-factor authentication on central-government sites ⁸² — and the recurring RENAPER incidents ⁸³. Confidence is capped at Low because the evidence subset (14 items) falls below the sufficiency threshold and the corpus entirely lacks market-share data. It does not descend to Level 1 because CERT.ar exists in a constitutive instrument and registers incidents, and a substantive strategy is published and under active reform. Argentina sits at the lower edge of the Brazilian band (Level 2–3) and well below Russia (Level 4, with Kaspersky and the sovereign GOST cryptography). An ascent to Level 3 would require a public trace of major incidents with response times, a domestic security-product industry of measurable share, and a sovereign cryptographic regime applied in critical infrastructure.

Intra-dimensional analysis. The four indicators of Dimension 2 share a single bottleneck: the Dependency Penalty. Argentina recognises every dependence — it has Compre Argentino, Public Software, a thriving software industry, and a CERT — but in each layer the real capacity is assembly (2.1), foreign upstream (2.2), vertical strength without horizontal substitution (2.3) or architecture without effectiveness (2.4). The symptomatic tension is that the dimension’s most visible asset — the software industry — feeds service exports to foreign clients and champions that are acquired by US firms, so that engineering capacity does not translate into infrastructure autonomy. The uniformity of Level 2 is no coincidence: it is the hallmark of an economy that is a net integrator of every layer of the digital stack.

Dimension 3 — Digital Governance Independence

Dimension 3 asks whether the country governs its digital space or is governed within others’ rules, through legislative capacity, enforcement, and leadership in international rules. The institutional landscape includes the Congreso de la Nación (National Congress), the AAIP, and the competition authority (CNDC, succeeded by the National Competition Authority), IRAM in technical standards and the Cancillería (Ministry of Foreign Affairs) in multilateral fora. The dimension finding (average 2.00) combines two high-confidence profiles — legislative capacity and enforcement, both capped at Level 2 by a subtractive trajectory and symbolic enforcement — with two low-confidence profiles owing to insufficient evidence (3.3 and 3.4), both also at Level 2: international participation without leadership.

3.3.1 Indicator 3.1 — Legislative capacity in digital matters

Rating: 2 — Aware. Confidence: High.

Argentina possesses a foundational but dated digital legal corpus: a cybercrime law (Law No. 26388, 2008), electronic signature (Law No. 25506, 2001), grooming (Law No. 26904, 2013), a statute on internet-provider content (Law No. 25690, 2002) and the telecommunications framework (Law No. 27078, 2014) ⁸⁴, alongside a data-protection statute (Law No. 25326, 2000) never substantively modernised. On a static reading, this corpus sits at the high end of Level 2: the foundational statutes are pre-2015 in substance, the frontier domains (artificial intelligence, platforms, statutory cybersecurity) are governed by executive instruments or non-binding guidance ⁸⁵, and the modernisation bills remain pending without enactment ⁸⁶.

Decisively, the trajectory is subtractive. The Ministry of Deregulation accumulates 543 measures that modify or eliminate 2,519 norms across 15,144 articles — verified textually against the official source ⁸⁷ — article 15 of Law No. 27078 was repealed by decree in April 2024 ⁸⁸ and the national artificial-intelligence strategy was institutionally discontinued ⁸⁹. By the indicator’s trajectory clause, a corpus under active dismantling cannot exceed Level 2 regardless of its historical base: this is the binding constraint, and it confirms Level 2 regardless of the static result of the decision tree. The indicator does not reach Level 3 because that would require at least one enacted — not proposed — frontier statute plus a trace of legislative modernisation in the last five years, conditions that the deliberate deregulatory trajectory (Argentina articulated as a ‘low-regulation hub’ for AI) ⁹⁰ renders unattainable. It does not descend to Level 1 because a genuine digital corpus is in force and the Budapest Convention and Convention 108+ have been incorporated. Against Brazil (Level 3, with the operative LGPD and an AI bill advancing), Argentina remains a level below precisely because its frontier bills do not advance to enactment and its existing corpus is being dismantled. The convergence with the BRICS 2025 rating (also Level 2) validates the reading. An ascent to Level 3 would require the enactment of a frontier statute (AI or platforms) and the cessation of the subtractive trajectory.

3.3.2 Indicator 3.2 — Enforcement capacity

Rating: 2 — Aware. Confidence: High.

Argentina has operative digital regulators with statutory sanctioning powers — the AAIP in data protection, the CNDC (replaced by the National Competition Authority on 17 November 2025) in competition, and the BCRA in financial cybersecurity — but enforcement is symbolic rather than consequential. The AAIP’s statutory fine scale runs from ARS 1,000 to ARS 100,000 (~USD 88–100 at its maximum) ⁹¹, orders of magnitude below the GDPR’s 4%-of-revenue standard; the largest sanction against a Big Tech firm ever imposed is the one against Google, for ARS 280,000 aggregated (~USD 215, 2020), equivalent to around 0.0001% of the offender’s revenue ⁹². The competition authority issued 89 merger-control decisions and a single procedural fine for late notification in 2024, with zero anticompetitive-conduct sanctions ⁹³.

The binding fact is the closure of the flagship case: the CNDC shelved the abuse-of-dominant-position investigation against WhatsApp and Meta on 2 July 2025 without a final sanction, after four years and two rounds of precautionary measures confirmed judicially ⁹⁴. A regulator that opened, litigated, and then abandoned its emblematic case against a foreign Big Tech firm without extracting any concession exemplifies the Level 2 ceiling. The indicator does not reach Level 3 because that requires at least one concluded enforcement action that produces documented behaviour change among the regulated, together with non-testimonial sanctions proportional to the infraction: the behaviour-change test is mandatory and fails resoundingly. It does not descend to Level 1 because the AAIP has concluded real monetary sanctions — even against a Big Tech firm — and the competition authority processes merger control at volume with professional staff. Argentina remains below Brazil (Level 2–3), whose CADE demonstrated the capacity to extract compliance even from foreign platforms (the X/Twitter episode), and well below China (Level 5, with the structural decisions against Alibaba and Didi). An ascent to Level 3 would require a concluded sanction with verifiable behaviour change and the updating of the fine cap to deterrent magnitudes.

3.3.3 Indicator 3.3 — Leadership in international technical rules

Rating: 2 — Aware. Confidence: Low.

Argentina is an active participant in international technical-standards bodies, with a genuine contribution but without leadership. It maintains its level of access through the general standards ecosystem — IRAM is the sole national representative before ISO and, together with CEA/AEA, before IEC, as well as a representative in the regional bodies COPANT and AMN ⁹⁵ — and the country rejoined IEC in 2001. The strongest and verified contribution is that of Verónica Marinelli, of IRAM’s SC 27 mirror committee, as one of three editors of the adopted international standard ISO/IEC 27002:2022 ⁹⁶; alongside her, editors from a single institution (UTN’s Santa Fe Regional Faculty) act as editors of at least five Y-series Recommendation projects of the ITU-T’s Study Group 20 ⁹⁷.

But the concentration in a single institution and a single body is exactly what the criterion caps at Level 2: an isolated editor/rapporteur contribution — the lowest leadership level — does not cross the frontier to Level 3, which requires multi-body adoptions and a handful of leadership roles. The chair leadership level is verifiably absent: there is no Argentine vice-chair in the 2025–2028 leadership roster of the ITU-T’s Study Group 20 ⁹⁸, nor any Argentine organisation among the 329 members of the W3C ⁹⁹ — Argentina is a rule-taker in web protocols. The regional engagement (chair of CITEL’s Steering Committee in 2018) is real but, by Pitfall #5, regional harmonisation is categorically distinct from global rule-making and does not raise the global rating ¹⁰⁰. Confidence is capped at Low because the corpus lost around ten candidate items to verification failures and ended up at 11 items, below the sufficiency threshold. Argentina groups with the Brazilian anchor (Level 2): both are regional leaders with limited global presence. An ascent to Level 3 would require contributions adopted in multiple bodies by multiple institutions, plus leadership roles above the editor level.

3.3.4 Indicator 3.4 — Leadership in international conduct rules

Rating: 2 — Aware. Confidence: Low.

Argentina is a recurrent and substantive participant in the main arenas of international digital governance — the UN Open-Ended Working Group on ICT security, the Ad Hoc Committee that produced the Convention on Cybercrime, the Council of Europe’s Convention 108 system and regional internet-governance fora — but the evidence shows participation without independent agenda-setting. The strongest signal, the election of Beatriz Anchorena to chair the Council of Europe’s Convention 108 Committee ¹⁰¹, is a genuine credential of treaty engagement, but it is a rotating institutional chairmanship over an instrument anchored in the European Union that advances the Council of Europe’s own work programme, not a doctrine of Argentine origin adopted by other States. The interventions in the Working Group were carried out explicitly ‘without co-sponsoring substantive thematic resolutions’ ¹⁰².

Decisively, a domestic–international coherence failure undermines any claim to independent normative leadership: a data-protection bill with extraterritorial scope and anti-CLOUD Act provisions (1948-D-2025) runs contemporaneously with the Framework Agreement that grants the United States data-adequacy status ¹⁰³ — Argentina advocates data sovereignty while conceding it. The indicator does not reach Level 3 because that requires co-sponsorship of a substantive instrument with documented impact on the final text, visible independent positioning, and domestic–international coherence, and the surviving evidence satisfies none of these. As to the Argentine withdrawal from the Pact for the Future and the Global Digital Compact, it is treated qualitatively as dissent-without-alternative — Level 2 positioning, not Level 3 leadership — without reproducing the specific date or cohort, whose evidentiary carrier was removed by the verifier. It does not descend to Level 1 because Argentina is an active participant of substance, with recurrent delegations and a treaty-committee chairmanship. Against Brazil (Level 3, with NETmundial 2014 and sustained leadership in the IGF), Argentina remains just below, at the upper edge of Level 2, for lacking a flagship convened process and for exhibiting the incoherence that Brazil does not show. An ascent to Level 3 would require a co-sponsored instrument with documented influence and the resolution of the domestic–international incoherence.

Intra-dimensional analysis. Dimension 3 exposes a coherent internal hierarchy: legislative capacity (3.1) exceeds enforcement (3.2) in its historical base, but both are capped at Level 2 for distinct reasons — a subtractive trajectory in 3.1, the absence of deterrent effect in 3.2 — satisfying the cross-cutting consistency rule that requires enforcement ≤ legislation. International leadership (3.3 and 3.4) remains at participation without influence, consistent with the rule that requires influence in international rules ≤ domestic R&D capacity. The common bottleneck is that Argentina possesses real institutions that operate within others’ rules: it legislates but dismantles, it enforces but does not deter, it participates but does not lead. The domestic–international incoherence of 3.4 — advocating data sovereignty while conceding US adequacy — is the dimension’s sharpest cross-cutting symptom.

Dimension 4 — Digital Capacity Independence

Dimension 4 analyses whether the country has the human, scientific, and industrial capacity to sustain digital sovereignty across generations, along four indicators: frontier research, university talent, industrial engineering, and strategic alignment. The institutional landscape includes CONICET, the free public university system, the MinCyT/Secretariat of Science and Technology, and the software industry. The dimension finding (average 2.25) makes it Argentina’s highest, sustained entirely by industrial engineering (4.3, the dimension’s only Level 3); the other three indicators remain at Level 2, cut across by the fiscal collapse of the period.

3.4.1 Indicator 4.1 — Frontier-technology research

Rating: 2 — Aware. Confidence: Medium.

Argentina possesses genuine institutional infrastructure for frontier digital research: CONICET, the first governmental scientific institution in Latin America, and the twentieth in the world among governmental research bodies ¹⁰⁴; a CNEA-CONICET superconducting-qubit programme (QUANTEC) ¹⁰⁵; a sovereign high-performance computing facility (Clementina XXI, 15.3 PFLOPs, ranked 82nd in the TOP500) ¹⁰⁶; and budgeted quantum-infrastructure calls ¹⁰⁷. What it does not possess is the binding condition that Level 3 requires: sustained presence in top-tier global venues (NeurIPS, ICML, ICLR) at a non-trivial volume over a five-year window. Argentina’s aggregate output at these venues is documented as low and dependent on foreign co-authorship ¹⁰⁸.

The existing artefacts are below the ‘more than a prototype’ threshold: QUANTEC targets 4–6 qubits, with coherent manipulation still a goal for 2025 ¹⁰⁹, and Clementina XXI is an acquired Lenovo cluster, a sovereign capacity but not a domestically designed artefact. The decisive modifier is a verified downward fiscal trajectory: R&D investment fell to 0.216% of GDP in 2024, a real decline of 31.3% and 55.4% below the 0.39% mandated by Law No. 27614 for that fiscal year ¹¹⁰, with that law’s progressivity articles suspended ¹¹¹. By Pitfall #3, the rating reflects the downward trajectory, not the inherited institutional stock. It does not descend to Level 1 because CONICET operates a substantive research community and there are funded programmes and sovereign HPC. Argentina remains a level below Brazil (Level 3), whose differential is a longer trajectory of sustained investment; the 2024 fiscal collapse sharpens the gap rather than narrowing it. An ascent to Level 3 would require sustained presence in top-tier venues and a funding trajectory that is at least stable.

3.4.2 Indicator 4.2 — University talent development

Rating: 2 — Aware. Confidence: Low.

Argentina operates a free and recognised university system that produces digital-technology graduates at a measurable but modest and declining volume, has no university among the world’s top 500 in computer science, and loses its most capable graduates to foreign labour markets through a structurally selective brain drain. The production system exists; the retention system has failed. The base is real: close to 6,500 engineering graduates per year nationally, with UTN producing 42.75% ¹¹², and specialised postgraduate programmes in data science, and artificial intelligence emerging in several universities ¹¹³. A State policy for repatriating the diaspora (RAICES) and pipeline programmes (Sadosky, Numéricas) evidence recognition of talent as a strategic asset ¹¹⁴.

But the indicator does not reach Level 3 for four reasons. No Argentine university appears in the global top 500 in computer science (UBA at 555th) ¹¹⁵. The production of computer-science graduates is small and decreasing (a decline of 23.68% in the 2010–2019 decade). The brain drain is the dominant structural pattern: Argentina is among the 30 largest emitters of highly qualified talent, with close to 30,000 scientists and engineers in the United States alone, and the service-export model rents domestic talent to foreign clients via remote work ¹¹⁶. And the fiscal trajectory collapses: science-and-technology investment around 0.2% of GDP (the level of the 2002 crisis) and CONICET doctoral scholarships cut by around 30% in 2024 ¹¹⁷. Confidence is capped at Low because the emigration rate specific to computer-science doctorates cannot be substantiated with Tier-1 sources and is sealed as a gap. It does not descend to Level 1 because the pipeline is operative and measurable. Argentina remains a level below Brazil (Level 3), whose sustained investment is the multi-year differential. An ascent to Level 3 would require a university in the top 200 in a digital discipline and the control of brain drain below the operational threshold.

3.4.3 Indicator 4.3 — Industrial engineering capacity

Rating: 3 — Developing. Confidence: High.

Argentina sits firmly at Level 3 in industrial engineering capacity. The industry has substantive scale: software-industry revenue reached USD 22,221 million in 2024 (+13.1% year-on-year) with 158,179 registered jobs ¹¹⁸, and more than 140,000 formal sector employees exceed the automotive, oil, and mining industries combined ¹¹⁹. Software and IT-service exports reached a record USD 2,674 million in 2024 and the knowledge-based-services export complex was USD 8,047 million in 2022 ¹²⁰, placing Argentina second in Latin America among software exporters. At least one subsector is internationally competitive: fintech, where three domestically headquartered digital banks (Ualá, Brubank, Naranja X) retain 88% of the digital-banking market with 16.44 million clients ¹²¹, and where the 2024 venture-capital rounds were led by fintech ¹²². This comfortably exceeds Level 2: it is a real industry of products and services with domestically headquartered firms of national consequence. Argentina has produced more than a dozen unicorn-scale companies over the past decade, of which Auth0 — built by Argentine engineers and sold to the US firm Okta for some USD 6,500 million in 2021 — is the emblematic surviving case in the verified corpus ¹²³.

But the configuration falls decidedly short of Level 4 owing to five verified structural limits. First, the export base is skewed towards services with a thin product layer: in the 2022 mix, software was 33%, professional services 30%, architecture and engineering 22%, and R&D only 8% ¹²⁴, and the sector ‘tends more towards development than research’ ¹²⁵. Second, the dominant exit pattern is foreign acquisition at the growth stage, not independent scaling: the Auth0→Okta case is the canonical pattern of the migration of intellectual property and decision rights abroad ¹²⁶. Third, Argentina loses global market share — from 0.6% of world software sales in 2011 to 0.3% in 2021 ¹²⁷. Fourth, the capital base is fragile: venture capital fell from a peak of USD 1,337 million in 2021 to USD 412 million in 2024 ¹²⁸. Fifth, there is no autonomy of the foundational layer: the largest announced technology investment is the OpenAI/Sur Energy data centre under the RIGI — foreign hyperscaler capacity settled in the territory, not domestic cloud or foundational-model capacity ¹²⁹. By Principle #4, an industry built on foreign cloud, models and design tools is engineering capacity deployed on others’ rails. Argentina sits on a par with Russia and Brazil (both Level 3) and well below China (Level 5); the differential with Brazil is precisely the Argentine capital fragility. An ascent to Level 4 would require service–product parity, autonomy of the foundational layer, a domestic venture-capital ecosystem at scale, and independent scaling without foreign-acquisition pressure.

3.4.4 Indicator 4.4 — Alignment with national strategy

Rating: 2 — Aware. Confidence: Medium.

Argentina possesses an abundant and sophisticated catalogue of digital and science-and-technology strategy instruments at the highest level of government: the National Science, Technology, and Innovation Plan 2030 enacted by Law No. 27738 ¹³⁰, the Digital Agenda 2030 by decree ¹³¹, the Strategic Guidelines 2025–2027 ¹³², a National Artificial Intelligence Plan, and an inter-ministerial AI roundtable spanning ten jurisdictions ¹³³. On the axis of the strategy document, the country is comprehensive. But on each execution axis that the indicator measures — budget alignment, statutorily funded coordination, indicator tracking, and continuity across administrations — the evidence shows symbolic strategy without execution.

The AI roundtable was created ‘without its own budget allocation’ ¹³⁴ and the National AI Plan lacks a budget; the only funding dedicated to AI is an external IDB loan of USD 35 million ¹³⁵, which by Principle #4 is execution driven by an external lender, not domestic strategic alignment. The science-and-technology budget function fell 50.6% in cumulative real terms across 2024–2026 ¹³⁶, the national budget has been carried over for two consecutive years ¹³⁷, and the flagship programmes are cancelled and relaunched cyclically — Conectar Igualdad was created, suspended, relaunched, restored, and defunded across four administrations ¹³⁸; the MinCyT was created, downgraded, restored, and downgraded again ¹³⁹. The governing strategic posture is active deregulation (DNU 70/2023, which repealed or modified 366 norms, including Law No. 27078 Argentina Digital) ¹⁴⁰ framed within an ambition to be a ‘low-regulation hub’ ¹⁴¹, which Principle #7 treats as hegemonic alignment, not as a sovereignty strategy. The indicator does not reach Level 3 because that requires a budget partially aligned with the priorities, coordination operating with funded decisions and sustained execution across at least one complete administrative mandate: the budget-alignment and continuity tests both fail. It does not descend to Level 1 because statutory-level strategies and a coordination architecture exist on paper. Argentina mirrors Brazil’s Level 2 through the same mechanism (political discontinuity) but in more acute form. An ascent to Level 3 would require a budget effectively aligned with strategy and the cessation of the cyclical-cancellation pattern.

Intra-dimensional analysis. Dimension 4 shows a broken synergy: Argentina has real human and industrial capacity (4.1, 4.2, 4.3) and abundant strategy (4.4), but the period’s fiscal collapse disarticulates them. Industrial engineering (4.3) sustains the dimension at Level 3, but its own ceiling — foundational-layer dependence, capital flight and talent flight — refers directly back to the weaknesses of 4.1 and 4.2. The explicit cross-cutting contradiction is that the AI strategy depends on a stock of university and research capacity that the same budget cuts are depleting, leaving the country, in the words of the evidence itself, as a data-centre operator rather than an AI innovator ¹⁴². The consistency rule that requires capacity ≤ infrastructure is satisfied: Dimension 4’s capacity does not exceed Dimension 2’s infrastructure, both anchored in structural dependence.

Chapter 4 — Cross-cutting Findings

4.1 Synthesis by dimension

The dimension-by-dimension reading reveals a country of uneven but internally coherent digital sovereignty. Dimension 1 (1.75) is the weakest: the historical legislative asset (1.1) does not prevent the simultaneous collapse of flow protection (1.3) and value capture (1.4) to Level 1. Dimension 2 (2.00) exhibits uniform awareness without substantive capacity: all four indicators at Level 2, cut across by the Dependency Penalty that the framework declares dominant for infrastructure. Dimension 3 (2.00) combines operative institutions with the absence of deterrent effect (3.2) and of international leadership (3.3, 3.4), its legislative capacity (3.1) capped by the subtractive trajectory. Dimension 4 (2.25), the highest, rests entirely on the software industry (4.3); its other three indicators remain at Level 2 owing to the period’s fiscal collapse.

4.2 Horizontal comparison: AR against CN, RU, BR

The DSI framework anchors horizontal consistency against China, Russia, and Brazil. Across the set of sixteen indicators, Argentina groups consistently with Brazil, sits systematically below Russia, and is at a wide structural distance from China. China operates as the ceiling in every dimension — an exporting and expanding legislative corpus, a domestic hardware and software ecosystem, international agenda-setting, and global industrial champions — sitting around Level 5 in most indicators. Russia occupies a Level 3–4 band, with sanctions-forced substitution that produced domestic foundational layers (Astra Linux, Yandex, GOST cryptography) that Argentina does not possess. Brazil is the closest comparator: both countries share transplanted data-protection regimes, Level 3 software and fintech industries, hyperscaler dependence in infrastructure, and regional participation without global leadership.

The key distinction between Argentina and Brazil is instructive. Brazil surpasses Argentina in three precise indicators: it has, in TOTVS, a domestic ERP provider with horizontal share (2.3), it advances an AI bill towards enactment (3.1) and it exhibits enforcement capacity that extracts compliance from foreign platforms (3.2). Argentina, in turn, retains an asset that Brazil does not have in the same form: the 2003 European adequacy and an older data-protection regime (1.1). But the decisive divergence is direction: where Brazil builds, Argentina — in the assessment period — dismantles. That directionality is what, via the trajectory clause (3.1) and Principle #7 (1.3), separates Argentina from the Level 3 band that Brazil reaches in several indicators.

Table 4.1. Approximate comparative positioning by dimension.

4.3 The computation of the sovereignty score

The digital sovereignty score is computed as the arithmetic mean of the sixteen indicator ratings on the 1–5 scale; the total is the sum of the sixteen ratings, with a maximum of 80. Argentina totals 32 out of 80, which yields a score of 2.00 / 5.0. The computation is transparent and consistent with the dimension averages: (1.75 + 2.00 + 2.00 + 2.25) / 4 = 2.00. The dimension that drags the score down most is Dimension 1 (data ownership), owing to its two indicators at Level 1; the one that most sustains it is Dimension 4, owing to the only Level 3 of industrial merit (4.3). The cross-cutting consistency check is satisfied on all its axes: enforcement (3.2, Level 2) does not exceed legislation (3.1, Level 2); international influence (3.3–3.4, Level 2) does not exceed R&D capacity (4.1, Level 2); and capacity (Dimension 4) is aligned with infrastructure (Dimension 2). No anomalous inversions are detected. The score of 2.00 places Argentina at the threshold between awareness and development: the country systematically recognises its sovereignty deficits but, except in two indicators, has not crossed over into the effective construction of independence.

4.4 The Milei pattern: active deregulation as a ceiling factor

The most significant cross-cutting finding is that the Milei administration’s trajectory operates as a ceiling factor that manifests independently in at least four indicators, through distinct methodological mechanisms. In indicator 1.3, the adequacy commitment with the United States activates the Principle #7 override and collapses the rating to Level 1 ¹⁴³. In indicator 3.1, the trajectory clause caps legislative capacity at Level 2 owing to the Ministry of Deregulation’s quantified normative dismantling — 543 measures over 2,519 norms ¹⁴⁴. In indicators 4.1 and 4.4, the fiscal collapse — a real cut of 50.6% in the science-and-technology function and the suspension of the Law No. 27614 financing law ¹⁴⁵ — invalidates the investment trajectory that Level 3 requires. The methodological distinction is important: the framework does not penalise the Argentine macroeconomic situation (that would be context, which by Principle #6 explains but does not excuse), but the active policy of capacity subtraction. Deregulation is not a circumstance Argentina suffers; it is a government decision that the index registers as direction.

4.5 The Argentine paradox

The second cross-cutting finding is the paradox between the regulatory sophistication of the data point and the collapse of its sovereignty in the flow. Argentina possesses the most mature data-protection framework in Latin America — the only one with European Union adequacy since 2003 ¹⁴⁶, reaffirmed in 2024 ¹⁴⁷ — which earns it the only Level 3 of Dimension 1 on its own legislative merit (1.1). But that same country, in the same period, commits to the recognition of the United States as an adequate jurisdiction, bringing the protection of cross-border flows down to Level 1 (1.3). The contrast is not accidental: the Law No. 25326 regime was designed looking towards Europe — it is a transplant of the European standard — and adequacy flows towards Argentina (inbound recognition), while the new US adequacy flows from Argentina towards the hegemonic jurisdiction of platforms (outbound cession). Argentina knows how to write European-class data laws, but at the decisive moment it cedes sovereignty over the outflow of those data to where the platforms that exploit them reside. The paradox extends to 1.4: the country that best legislates the data point is also the one that least exploits it economically.

4.6 Structural and multi-actor constraints

The multi-actor analysis confirms these patterns. The government of the period prioritises deregulation and the attraction of foreign investment (RIGI) over the construction of sovereign capacity. The large foreign platforms capture cloud growth (AWS, OpenAI/Sur Energy) ¹⁴⁸ and acquire domestic champions (Auth0→Okta) ¹⁴⁹, in a pattern that the framework reads as corporate capture and dependency penalty. Civil society and academia sustain real capacity — civic-tech, CONICET, the free university system — but without budget or retention: the most capable talent emigrates ¹⁵⁰ and scientific investment collapses ¹⁵¹. The dominant geopolitical constraint is the realignment with the United States, which materialises the sovereignty cost of Principle #7. By Principle #6, these circumstances explain the pattern but do not raise any rating.

4.7 Prospective synthesis

Should the deregulation trajectory be sustained through 2027 — fiscal cuts in science and technology, deepening of US adequacy, absence of legislative modernisation — Argentina’s average score would tend to decline from the current 2.00, driven by additional erosion in Dimensions 1 and 4 and by the downward pressure already noted in 2.1 (collapse of the Fuegian assembly base) and 4.1 (depletion of the research stock). Chapter 5 develops the scenarios and actions that could reverse this direction.

Chapter 5 — Conclusions and Recommendations

5.1 Five principal conclusions

First: Argentina is a country of mid-to-low-grade digital sovereignty (2.00 / 5.0) without a single competent indicator. The matrix records no Level 4 or 5; its two high points (1.1 and 4.3) reach Level 3, and two low points (1.3 and 1.4) fall to Level 1. The country systematically recognises its deficits — it is ‘aware’ in ten of sixteen indicators — but awareness has not translated into effective capacity ¹⁵².

Second: data ownership is the central fracture. With an average of 1.75, Dimension 1 concentrates the Argentine weakness, not for absence of a legal framework but for its paradoxical use: a sophisticated protection regime (1.1) that does not retain data territorially (1.2), cedes them to the hegemonic jurisdiction (1.3), and does not capture their value (1.4) ¹⁵³.

Third: structural dependence is cross-cutting and concentrated in infrastructure and capacity. Dimensions 2 and 4 show that Argentine engineering capacity — real and of international scale — is deployed on others’ rails: hardware assembled but not manufactured, software with foreign upstream, an industry that exports services, and exports its talent ¹⁵⁴.

Fourth: the deregulation of the period operates as a ceiling factor, not as context. US adequacy (1.3), normative dismantling (3.1), and fiscal collapse (4.1, 4.4) are active policies that the framework registers as a downward direction, not as mitigating circumstances ¹⁵⁵.

Fifth: Argentina groups with Brazil, separated by direction. The closest regional comparator surpasses Argentina in the indicators where it builds (domestic ERP, AI law, effective enforcement), while Argentina retains its historical advantage only where it inherited capacity (1.1). Directionality — construction versus dismantling — is the decisive separator ¹⁵⁶.

5.2 Five recommendations

1. To the Congreso de la Nación: enact the data-protection reform within the term of the next ordinary session. The modernising bill (successor to Message 87/2023, lost for lack of consideration) must enter into force with regulated breach notification, impact assessments, the data-protection officer, and, non-negotiably, a fine cap no lower than 2% of the offender’s annual revenue, to restore the deterrence that the USD 70–100 scale has lost ¹⁵⁷. Without updating the cap, indicators 1.1 and 3.2 will remain capped at Level 3 and Level 2 respectively.
2. To the Executive and the Cancillería: condition the Framework Agreement’s US adequacy commitment on equivalent protection guarantees. In the short term, before the operational implementation of the Agreement on Reciprocal Trade and Investment, the Executive should incorporate safeguards — redress mechanisms for data subjects and limitation of extraterritorial access — that mitigate the sovereignty cost that Principle #7 penalises ¹⁵⁸. Without it, indicator 1.3 will remain collapsed at Level 1.
3. To the Ministry of Economy and the Secretariat of Innovation, Science and Technology: align budget with strategy and restore the path of Law No. 27614. Within the term of the next national budget (ending the two-year carryover), the inter-ministerial AI roundtable, and the National AI Plan should be endowed with their own budget allocation, and the suspended progressivity articles of Law No. 27614 should be reactivated ¹⁵⁹. Indicator 4.4 will not exceed Level 2 while the strategy lacks a living budget.
4. To the private sector and the promotion agencies: orient the Knowledge Economy incentives towards horizontal substitution and the foundational layer. In the medium term, the tax benefits of Law No. 27506 should be conditioned on the generation of product — not only export services — and on the retention of intellectual property in the country, to reverse the foreign-acquisition pattern exemplified by the Auth0→Okta case ¹⁶⁰. This is a prerequisite for indicator 4.3 to move towards Level 4.
5. To academia and CONICET: institutionalise talent retention and production in frontier venues. In the medium and long term, the restoration of doctoral scholarships and the creation of retention incentives and of incentives to publish in top-tier venues — linked to the already-existing sovereign infrastructure (Clementina XXI, QUANTEC) — are a condition for moving indicators 4.1 and 4.2 ¹⁶¹.

5.3 Future trajectory: three scenarios

Favourable scenario. If Congress enacted the data reform with deterrent caps, the Executive introduced safeguards to US adequacy and the budget restored the science-and-technology path, Argentina could recover Level 4 in 1.1 and Level 3 in 1.3 and 3.2, raising the average score towards 2.3–2.5 over a horizon of three to five years.

Continuity scenario. Should current trends hold — strategy without budget, stalled legislative modernisation, US adequacy in force — the score would hold around 2.00, with the industrial strength (4.3) compensating for the gradual erosion in other areas.

Unfavourable scenario. If deregulation were to deepen — collapse of the Fuegian assembly base, depletion of the research stock, statutory codification of US adequacy — the score would decline towards 1.7–1.8, with indicator 2.1 sliding to Level 1 and downward pressure across the whole of Dimension 4.

5.4 Future research questions

This report leaves open five lines that the evidence and time limitations did not allow to resolve. First: what is the emigration rate specific to doctorates in computer science and artificial intelligence, today sealed as a gap for lack of a Tier-1 source? Second: what effective share of public software spending corresponds to providers of national origin, a figure that neither COMPR.AR nor the AGN segments? Third: what are the rates of judicial confirmation and of effective collection of the administrative fines of the AAIP and the competition authority? Fourth: what operational impact will the new architecture of Decree 941/2025 (National Cybersecurity Centre, Federal Cyberintelligence Agency) have once it produces measurable activity? Fifth: how will Argentine participation in international technical standards evolve if the URL_BLOCKED items from ITU and 3GPP documentary databases, which the verifier could not access, are recovered?

References (by Dimension)

The references are grouped by dimension and list, in ascending order, each evidence identifier actually cited in the body of the report. The complete library is kept in evidence_base_UNGS_2026.json. Corpus access date: 2026-05-29.

Dimension 1 — Data ownership

• ¹⁶² Law No. 25326 on Personal Data Protection (art. 12; 2000 fine scale). Regulation, Tier-1. Source: infoleg.gob.ar.
• ¹⁶³ Decree 1558/2001, implementing Law No. 25326 (in force, designates the enforcement authority). Regulation, Tier-1. Source: infoleg.gob.ar.
• ¹⁶⁴ Agencia de Acceso a la Información Pública (AAIP) — regulator with the National Directorate for the Protection of Personal Data. Regulation, Tier-1. Source: argentina.gob.ar/aaip.
• ¹⁶⁵ Data-protection modernisation bill (Message 87/2023), loss of parliamentary status. Policy, Tier-1. Source: HCDN.
• ¹⁶⁶ AAIP Resolution 47/2018 — security measures without geographic restriction. Regulation, Tier-1. Source: argentina.gob.ar/aaip.
• ¹⁶⁷ Official AAIP guidance on adequacy; United States absent; EU adequacy reaffirmed January 2024. Report, Tier-1. Source: argentina.gob.ar/aaip.
• ¹⁶⁸ Law No. 27699 — ratification of Convention 108+ (2022). Regulation, Tier-1. Source: infoleg.gob.ar.
• ¹⁶⁹ USTR Fact Sheet on the US–Argentina Framework Agreement (13 Nov. 2025). Policy, Tier-1. Source: ustr.gov.
• ¹⁷⁰ AAIP 2022 annual figures (491 case files, 52 sanctions, ~USD 30,621). Data, Tier-1. Source: argentina.gob.ar/aaip.
• ¹⁷¹ Gap — absence of breach notification, DPIA, DPO, and extraterritorial clause in the text in force. Gap, Tier-1.
• ¹⁷² BCRA Communication A 7724 — prudential regime for third-party control. Regulation, Tier-1. Source: bcra.gob.ar.
• ¹⁷³ ARSAT’s Tier III data centre — residual sovereign capacity. Report, Tier-1. Source: arsat.com.ar.
• ¹⁷⁴ Executive reform bill (2023) — preserves the transfer model. Policy, Tier-2.
• ¹⁷⁵ Reform bill 1948-D-2025 — continuity of the non-localising approach. Policy, Tier-1. Source: HCDN.
• ¹⁷⁶ Installed domestic IT capacity (~32 MW). Data, Tier-3.
• ¹⁷⁷ OpenAI/Sur Energy 500 MW project under the RIGI. CaseStudy, Tier-2.
• ¹⁷⁸ AWS Local Zone Buenos Aires — presence under foreign jurisdiction. Report, Tier-2.
• ¹⁷⁹ RIGI/Super RIGI — incentives for foreign-controlled infrastructure. Policy, Tier-1.
• ¹⁸⁰ Partial privatisation (49%) of ARSAT. Policy, Tier-1.
• ¹⁸¹ Cross-border personal-data regime (Law No. 25326 art. 12) based on transfers. Regulation, Tier-1.
• ¹⁸² BCRA Communication A 8401 — update of the sectoral regime. Regulation, Tier-1.
• ¹⁸³ Decree 1558/2001 — the regulator’s adequacy-assessment power. Regulation, Tier-1.
• ¹⁸⁴ Disposition 60-E/2016 — model contractual clauses. Regulation, Tier-1.
• ¹⁸⁵ AAIP Resolution 34/2019 — adequacy list; US absent. Regulation, Tier-1.
• ¹⁸⁶ AAIP Resolution 198/2023 — model transfer clauses. Regulation, Tier-1.
• ¹⁸⁷ Joint Statement of the US–Argentina Framework (13 Nov. 2025). Policy, Tier-3. Source: whitehouse.gov.
• ¹⁸⁸ Signing of the Agreement on Reciprocal Trade and Investment (5 Feb. 2026). Policy, Tier-1. Source: cancilleria.gob.ar.
• ¹⁸⁹ EU adequacy maintained since 2003, revalidated January 2024. Policy, Tier-2.
• ¹⁹⁰ Historical operation of US Big Tech via contractual clauses. Report, Tier-2.
• ¹⁹¹ Gap — absence of transfer-enforcement statistics. Gap, Tier-1.
• ¹⁹² Decree 117/2016 — open-data programme on a transparency logic. Regulation, Tier-1.
• ¹⁹³ Law No. 27275 on access to public information. Regulation, Tier-1.
• ¹⁹⁴ Decree 780/2024 — restriction of the scope of active transparency. Regulation, Tier-1.
• ¹⁹⁵gob.ar portal — 1,235 datasets from 42 agencies. Data, Tier-1.
• ¹⁹⁶ PAIS tax — consumption tax repealed in December 2024. CaseStudy, Tier-1.
• ¹⁹⁷ CNDC — digital-markets study group without sanctioning power. CaseStudy, Tier-1.
• ¹⁹⁸ CNDC — eleven concentrated markets, none a digital-native platform. CaseStudy, Tier-1.
• ¹⁹⁹ Civic-tech ecosystem reusing public data. CaseStudy, Tier-2.
• ²⁰⁰ Gap — absence of an operative B2G mandate. Gap, Tier-1.
• ²⁰¹ Gap — absence of recognition of data as a factor of production. Gap, Tier-1.
• ²⁰² Gap — OECD Pillar One and Pillar Two not implemented. Gap, Tier-1.
• ²⁰³ Gap — absence of institutional data trusts. Gap, Tier-2.

Dimension 2 — Digital infrastructure

• ²⁰⁴ Compre Argentino (Law No. 27437) — public-procurement preference. Regulation, Tier-1.
• ²⁰⁵ Tierra del Fuego regime (Law No. 19640) — assembly (~93% of phones, AC, TV). Report, Tier-2.
• ²⁰⁶ Decree 333/2025 — mobile-phone tariff reduction 8%→0% by January 2026. Regulation, Tier-1.
• ²⁰⁷ RIGI — enclave model without a technology-transfer obligation. Policy, Tier-1.
• ²⁰⁸ Knowledge Economy (Law No. 27506) — benefits to software/services. Regulation, Tier-1.
• ²⁰⁹ Design-only semiconductor capacity (INTI/IDME), no wafer plant. Report, Tier-2.
• ²¹⁰ Mirgor Group — suspension of ~360 workers. CaseStudy, Tier-2.
• ²¹¹ CTI Guidelines 2025–2027 — omission of microelectronics/semiconductors. Policy, Tier-1.
• ²¹² Desktop OS share (Windows 81.64%; Linux 2.21%). Data, Tier-3.
• ²¹³ Mobile market >99.97% foreign. Data, Tier-3.
• ²¹⁴ State Huayra GNU/Linux distribution (EDUCAR). CaseStudy, Tier-2.
• ²¹⁵ Suspension of Conectar Igualdad / Educ.ar (‘construction mode’). CaseStudy, Tier-1.
• ²¹⁶ Decree 963/2024 — intervention of EDUC.AR S.E. Regulation, Tier-1.
• ²¹⁷ National Public Cloud (ARSAT) on open-source code. Report, Tier-1.
• ²¹⁸ ARSAT runs its NOC on Red Hat OpenShift (foreign control plane). CaseStudy, Tier-2.
• ²¹⁹ ONTI Public Software initiative. Policy, Tier-1.
• ²²⁰ ONTI Disposition 2/2019 — code sharing under open licences. Regulation, Tier-1.
• ²²¹ Santa Fe provincial free-software law (Law No. 12360). Regulation, Tier-2.
• ²²² Database market dominated by Oracle/SAP/Microsoft/IBM. Data, Tier-3.
• ²²³ Official argob repositories (43 repositories). Data, Tier-1.
• ²²⁴ Mi Argentina — state platform, 21 million registered users. CaseStudy, Tier-1.
• ²²⁵ Software/IT-services exports USD 2,674M (2024), 46.9% to the US. Data, Tier-3.
• ²²⁶ Law No. 27506 Knowledge Economy (promotion). Regulation, Tier-1.
• ²²⁷ Mercado Libre — national-download leader 2024 (11.7M). Data, Tier-3.
• ²²⁸ Mercado Pago — 50M+ active users LATAM, USD 8,600M revenue. Data, Tier-3.
• ²²⁹ Digital banking — 88% across three domestic neobanks. Data, Tier-3.
• ²³⁰ Foreign horizontal consumer categories (WhatsApp 93%, Instagram 86.7%, TikTok 58.4%). Data, Tier-3.
• ²³¹ Search — Google 94.01%. Data, Tier-3.
• ²³² Auth0→Okta acquisition. CaseStudy, Tier-3.
• ²³³ Gap — ERP/CRM without auditable domestic share. Gap, Tier-2.
• ²³⁴ar (Disposition 1/2021). Regulation, Tier-1.
• ²³⁵ar — 438 incidents in 2024 (61% state sector). Data, Tier-1.
• ²³⁶ Administrative Decision 641/2021 — minimum security requirements and 48h notification. Regulation, Tier-1.
• ²³⁷ Decree 941/2025 — National Cybersecurity Centre and Federal Cyberintelligence Agency. Regulation, Tier-1.
• ²³⁸ Leak of the national driver’s-licence database (~5.7–6M, 1.25 TB, April 2024). CaseStudy, Tier-2.
• ²³⁹ Defacement of Mi Argentina/SUBE (Dec. 2024), no 2FA on central sites. CaseStudy, Tier-2.
• ²⁴⁰ Recurring RENAPER incidents. CaseStudy, Tier-2.
• ²⁴¹ Gap — absence of sovereign cryptographic research and a security industry. Gap, Tier-1.
• ²⁴² Gap — absence of documented sovereign offensive/red-team capacity. Gap, Tier-1.

Dimension 3 — Digital governance

• ²⁴³ Law No. 26388 on cybercrime (2008). Regulation, Tier-1.
• ²⁴⁴ Law No. 25506 on electronic signature (2001). Regulation, Tier-1.
• ²⁴⁵ DNU 302/2024 — repeal of art. 15 of Law No. 27078. Regulation, Tier-1.
• ²⁴⁶ AI governance — non-binding AAIP programme. Policy, Tier-1.
• ²⁴⁷ ‘Low-regulation hub’ positioning for AI. Analysis, Tier-2.
• ²⁴⁸ Ministry of Deregulation — 543 measures / 2,519 norms / 15,144 articles (verified). Data, Tier-1.
• ²⁴⁹ Absence of a platform statute; intermediary-liability bill without enactment. Policy, Tier-2.
• ²⁵⁰ Cybersecurity by DNC dispositions, without a parliamentary statute. Regulation, Tier-1.
• ²⁵¹ Discontinuation of the national AI strategy. Policy, Tier-2.
• ²⁵² AAIP statutory fine scale (ARS 1,000–100,000). Regulation, Tier-1.
• ²⁵³ AAIP Resolution 126/2024 — tiering of sanctions. Regulation, Tier-1.
• ²⁵⁴ Shelving of the CNDC investigation against WhatsApp/Meta (2 Jul. 2025). CaseStudy, Tier-2.
• ²⁵⁵ Largest sanction against a Big Tech firm (Google, ARS 280,000, 2020). CaseStudy, Tier-2.
• ²⁵⁶ CNDC 2024 — 89 merger decisions, one procedural fine, zero conduct sanctions. Data, Tier-1.
• ²⁵⁷ IRAM — sole representative before ISO and (with CEA/AEA) IEC; COPANT/AMN. Report, Tier-2.
• ²⁵⁸ UTN Santa Fe — editors of five Y-series Recommendations of the ITU-T’s SG20. CaseStudy, Tier-2.
• ²⁵⁹ Verified absence of an Argentine vice-chair in the ITU-T’s SG20 (2025-2028). Data, Tier-2.
• ²⁶⁰ ISO/IEC 27002:2022 — Argentine editor (IRAM SC 27). CaseStudy, Tier-2.
• ²⁶¹ CITEL — chair of the Steering Committee (2018) and rapporteurships. CaseStudy, Tier-2.
• ²⁶² Verified absence of an Argentine organisation among the 329 members of the W3C. Data, Tier-2.
• ²⁶³ US–Argentina Framework Agreement with data adequacy. Policy, Tier-2.
• ²⁶⁴ Participation in the OEWG without co-sponsorship of substantive resolutions. Report, Tier-2.
• ²⁶⁵ Election of Beatriz Anchorena to the chairmanship of the Convention 108 Committee. CaseStudy, Tier-2.
• ²⁶⁶ Data bill with extraterritorial scope / anti-CLOUD Act (1948-D-2025). Policy, Tier-1.

Dimension 4 — Digital capacity

• ²⁶⁷ R&D investment 0.216% of GDP (2024), -31.3% real, 55.4% below Law No. 27614. Data, Tier-3.
• ²⁶⁸ Suspension of the progressivity articles of Law No. 27614. Regulation, Tier-1.
• ²⁶⁹ Clementina XXI — sovereign HPC, 15.3 PFLOPs, TOP500 #82. Data, Tier-2.
• ²⁷⁰ QUANTEC — CNEA-CONICET superconducting-qubit programme (4-6 qubits). Report, Tier-1.
• ²⁷¹ MinCyT quantum-infrastructure call. Policy, Tier-1.
• ²⁷² CONICET — #1 LatAm / #20 worldwide among governmental research bodies. Data, Tier-2.
• ²⁷³ Argentine output at top-tier venues: low, dependent on foreign co-authorship. Analysis, Tier-3.
• ²⁷⁴ CONICET doctoral scholarships cut ~30% (950 in 2024). Data, Tier-2.
• ²⁷⁵ S&T investment ~0.2% of GDP (2002 level). Data, Tier-2.
• ²⁷⁶ University pipeline (~6,500 engineering graduates/year; UTN 42.75%). Data, Tier-2.
• ²⁷⁷ Brain drain (top-30 emitter; ~30,000 scientists/engineers in the US). Data, Tier-2.
• ²⁷⁸ Remote-work model for foreign clients. Analysis, Tier-2.
• ²⁷⁹ RAICES — diaspora repatriation policy (Law No. 26421). Policy, Tier-1.
• ²⁸⁰ No Argentine university in the global CS top 500 (UBA #555). Data, Tier-3.
• ²⁸¹ Software industry USD 22,221M (2024), 158,179 jobs. Data, Tier-3.
• ²⁸² SBC mix 2022 (software 33%, services 30%, architecture 22%, R&D 8%); USD 8,047M. Data, Tier-2.
• ²⁸³ Global software share 0.6% (2011) → 0.3% (2021). Data, Tier-2.
• ²⁸⁴ Auth0 — acquired by Okta (~USD 6,500M, 2021). CaseStudy, Tier-3.
• ²⁸⁵ Venture capital USD 1,337M (2021) → USD 412M (2024). Data, Tier-2.
• ²⁸⁶ OpenAI/Sur Energy data centre under the RIGI. CaseStudy, Tier-2.
• ²⁸⁷ The sector tends more towards development than research. Analysis, Tier-2.
• ²⁸⁸ >140,000 formal software employees (exceeds automotive+oil+mining). Data, Tier-2.
• ²⁸⁹ Law No. 27738 — National CTI Plan 2030. Regulation, Tier-1. Source: boletinoficial.gob.ar.
• ²⁹⁰ Strategic Guidelines 2025–2027. Policy, Tier-1.
• ²⁹¹ Inter-ministerial AI roundtable ‘without its own budget allocation’. Regulation, Tier-1.
• ²⁹² IDB loan of USD 35M as the only funding dedicated to AI. Policy, Tier-2.
• ²⁹³ Digital Agenda 2030 (Decree 996/2018). Regulation, Tier-1.
• ²⁹⁴ DNU 70/2023 — economic emergency, 366 norms repealed/modified. Regulation, Tier-1.
• ²⁹⁵ MinCyT — create/downgrade/restore/downgrade cycle. Report, Tier-2.
• ²⁹⁶ S&T budget function -50.6% cumulative real 2024-2026. Data, Tier-1.
• ²⁹⁷ Conectar Igualdad — cancellation/relaunch across four administrations. CaseStudy, Tier-2.
• ²⁹⁸ ‘Low-regulation hub’ positioning. Policy, Tier-2.
• ²⁹⁹ Fundar — the ‘Achilles’ heel’ of the AI strategy (university stock without investment). Analysis, Tier-2.
• ³⁰⁰ Budget carryover for two consecutive years. Report, Tier-2.

Appendix A — Inventory of Cited Evidence

This report cites 120 distinct evidence identifiers from the UNGS_2026 corpus (out of a total of 241 integrated pieces of evidence). All cited identifiers exist and are traceable in evidence_base_UNGS_2026.json. The following table summarises the composition of the cited evidence by type and source tier; the detail by identifier appears in the preceding References section.

Table A.1. Composition of cited evidence by type (approximate count).

Table A.2. Distribution by source tier.

Table A.3. Count of integrated evidence by indicator (full corpus, from evidence_index_UNGS_2026.json).

Appendix B — Documented Gaps

The UNGS_2026 corpus records 31 gaps under a strict policy (gap_policy_strict): confirmed absences of evidence after three rounds of search in Tier-1 sources, treated as evidence of a structural absence. They are enumerated by dimension.

Dimension 1 (9 gaps)

• 1 — GDPR-era modernisation absent: no regulated breach notification, mandatory DPIA, DPO figure, or extraterritorial clause in the text in force.
• 2 — No localisation-enforcement actions (2024–2026): a structural absence, there is no general mandate to enforce.
• 3 — No quantitative cross-border transfer-enforcement statistics published by the AAIP.
• 4 (×6) — Antitrust over data monopolies non-existent; open-data outcome audit pending; operative B2G mandate absent; institutional data trusts absent; OECD Pillar One/Two not implemented; recognition of data as a factor of production absent.

Dimension 2 (9 gaps)

• 1 (×2) — No Tier-1 data on domestic-vs-imported hardware share in public procurement; no active strategic hardware-stockpile programme.
• 2 (×3) — No Tier-1 quantification of public spending on foreign proprietary-provider licences; no documented event of sanctions/export controls on foundational software; no domestic capacity in kernels, hypervisors, or cryptographic libraries at scale.
• 3 (×2) — No auditable ERP/CRM/SCM breakdown by provider origin; no official metrics of the domestic-vs-foreign share in public spending on application software.
• 4 (×2) — No sustained domestic academic cryptographic publication at top-tier venues; no state red-team programme with public outputs.

Dimension 3 (9 gaps)

• 1 — No binding national framework for coordinated vulnerability disclosure.
• 2 (×2) — No systematic data on judicial confirmation/revocation or decision times; no quantification of case-to-decision time or fine-collection rate.
• 3 (×3) — No surfaceable direct Argentine authorship of an IETF RFC; no declared Argentine SEPs or participation in post-quantum standardisation; no documented official technical positions on disputed standards (‘New IP’, AI security, quantum communication).
• 4 (×3) — Argentina absent from the AI-safety declarations (Bletchley, Seoul, Paris); absent from the 65 initial signatories of the Convention on Cybercrime; no signature of the China-led Global Data Security Initiative.

Dimension 4 (4 gaps)

• 1 — No Tier-1 source with systematic annual counts of Argentine publications at top-tier AI venues.
• 2 — No disaggregated brain-drain rate in computer-science/AI subfields or directional flow of postgraduate students.
• 3 — No quantitative comparison of gross value added per worker vs. the OECD (PDF not retrievable).
• 4 — No Tier-1 dashboard of digital KPIs in the Presidency, JGM, or Economy tracking the CTI Plan 2030.

Appendix C — Authorship, Cooperation and Traceability

This Argentina assessment is a product of academic cooperation developed under the Global South Academic Forum, with the following division of roles:

• Global South Insights (GSI) — DSI Assessment Team. Design and specification of the Digital Sovereignty Index methodological framework (the 4×4 analytical structure: four dimensions, sixteen indicators, a 1–5 scale, seven evidence types, seven principles, and the decision tree), coordination of the international assessment, and compilation of the annual global index.
• National University of General Sarmiento (UNGS) — University Programme on Digital Sovereignty. Implementation of the Argentina assessment as an associate university: evidence collection from national sources, local application of the assessment pipeline in accordance with the DSI specifications, and drafting of the analysis in Argentine academic Spanish.
• Global South Academic Forum (GSAF). Institutional framework of the cooperation.
• Artificial-intelligence assistance. Evidence processing, verification, and drafting were assisted by the Claude model (Anthropic) — Co-authored-by: Claude.

The application of the specifications to the Argentina assessment — what national evidence is sufficient, what evidentiary weight corresponds to each local regulatory action, and how the analysis is written in Spanish — constitutes UNGS’s substantive contribution to the global index, in line with the methodology established by GSI.

Corpus traceability. Run: UNGS_2026. Phase-gate results: Phase 1 (collection, 373 pieces of evidence) → verification (91 removed) → integration (41 duplicates removed, 241 integrated + 31 gaps) → Phase 4 (assessment, 16 indicators) → Phase 5 (generation of this report). The evidence base (evidence_base_UNGS_2026.json) and the indexes (evidence_index_UNGS_2026.json) constitute the audit trail; each [AR-EV-NNN] citation in the report is verifiable against a corpus URL. Report generation date: 29 May 2026. The 2025 assessment is preserved as a comparative baseline.

Note

¹ AR-EV-0035: ‘En enero de 2024 la Comisión Europea publicó el primer informe de revisión de las decisiones de…’ Policy. https://www.argentina.gob.ar/noticias/argentina-logro-la-nueva-adecuacion-por-parte-de-la-union-europea-para-el-flujo. Accessed 2026-05-29.

² AR-EV-0212: ‘Argentina’s software industry reached USD 22,221 million in revenue in 2024 (+13.1% interannual)…’ Data. https://www.itsitio.com/ar/software/software-argentino-bate-records/. Accessed 2026-05-29.; AR-EV-0101: ‘El segmento de digital banking argentino está concentrado en tres jugadores domésticos: Ualá (6…’ Data. https://fintechnews.am/fintech-argentina/52228/argentinas-top-3-digital-banks-capture-nearly-90-market-share/. Accessed 2026-05-29.

³ AR-EV-0076: ‘Huayra GNU/Linux es el sistema operativo libre desarrollado en EDUCAR Sociedad del Estado, descrito…’ CaseStudy. https://huayra.educar.gob.ar/. Accessed 2026-05-29.; AR-EV-0079: ‘ARSAT desplegó la etapa 1 de la Nube Pública Nacional el 1° de abril de 2021, construida sobre…’ CaseStudy. https://www.canal-ar.com.ar/29335-ARSAT-lanzo-la-Nube-Publica-Nacional-y-ofrece-infraestructura-y-servicios-a-demanda.html. Accessed 2026-05-29.; AR-EV-0093: ‘Mi Argentina, la plataforma de identidad digital ciudadana, supera los 21 millones de personas…’ CaseStudy. https://www.argentina.gob.ar/noticias/se-renueva-mi-argentina-con-mas-y-mejores-servicios-para-la-ciudadania. Accessed 2026-05-29.

⁴ AR-EV-0031: ‘El 13 de noviembre de 2025, los presidentes Donald J. Trump y Javier Milei suscribieron el Joint…’ Policy. https://www.whitehouse.gov/briefings-statements/2025/11/joint-statement-on-framework-for-a-united-states-argentina-agreement-on-reciprocal-trade-and-investment/. Accessed 2026-05-29.; AR-EV-0032: ‘El 5 de febrero de 2026 en Washington D.C., la República Argentina y los Estados Unidos…’ Policy. https://www.cancilleria.gob.ar/es/destacados/argentina-y-estados-unidos-firmaron-un-acuerdo-sobre-comercio-e-inversiones-reciprocos. Accessed 2026-05-29.

⁵ AR-EV-0054: ‘No se identifica fuente Tier-1 que documente la existencia de un mandato general operativo de…’ Gap. Accessed 2026-05-29.; AR-EV-0055: ‘No se identifica fuente Tier-1 que documente reconocimiento estatal explícito de los datos como…’ Gap. Accessed 2026-05-29.; AR-EV-0057: ‘No se identifica fuente Tier-1 que documente la implementación por parte de Argentina de mecanismos…’ Gap. Accessed 2026-05-29.

⁶ AR-EV-0064: ‘El Instituto Nacional de Tecnología Industrial (INTI), a través de su área de Micro y…’ CaseStudy. https://www.inti.gob.ar/areas/desarrollo-tecnologico-e-innovacion/areas-de-conocimiento/micro-y-nanotecnologias. Accessed 2026-05-29.; AR-EV-0080: ‘ARSAT desplegó Red Hat OpenShift AI para apoyar operaciones de centro de operaciones de red (NOC)…’ CaseStudy. https://www.redhat.com/en/success-stories/arsat. Accessed 2026-05-29.; AR-EV-0200: ‘Argentina is in the top 30 nations with highly-skilled emigrants according to OECD data, with…’ Data. https://www.untref.edu.ar/mundountref/argentina-se-convirtio-en-un-polo-de-emigracion. Accessed 2026-05-29.

⁷ AR-EV-0133: ‘The Ministerio de Desregulación y Transformación del Estado (created July 2024 under Federico…’ Data. https://www.argentina.gob.ar/desregulacion. Accessed 2026-05-29.

⁸ AR-EV-0235: ‘La Función Ciencia y Tecnología del Presupuesto Nacional cayó un 11,4% adicional en el primer…’ Data. https://ciicti.org/el-2026-llego-con-mas-recortes-presupuestarios-a-la-ciencia-argentina/. Accessed 2026-05-29.

⁹ AR-EV-0035: ‘En enero de 2024 la Comisión Europea publicó el primer informe de revisión de las decisiones de…’ Policy. https://www.argentina.gob.ar/noticias/argentina-logro-la-nueva-adecuacion-por-parte-de-la-union-europea-para-el-flujo. Accessed 2026-05-29.

¹⁰ AR-EV-0001: ‘El artículo 12 de la Law No. 25326 prohíbe la transferencia de datos personales de cualquier tipo con…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/60000-64999/64790/texact.htm. Accessed 2026-05-29.

¹¹ AR-EV-0002: ‘El Decree 1558/2001, promulgado el 29 de noviembre de 2001 y publicado en el Boletín Oficial el 3…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/verNorma.do?id=70368. Accessed 2026-05-29.

¹² AR-EV-0003: ‘La AAIP se autodefine en su sitio oficial como garante de la protección de datos personales y la…’ Regulation. https://www.argentina.gob.ar/aaip/datospersonales. Accessed 2026-05-29.

¹³ AR-EV-0005: ‘La AAIP aprobó por Resolución 47/2018 (publicada 25-07-2018, vigente) las Medidas de Seguridad…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/310000-314999/312662/norma.htm. Accessed 2026-05-29.; AR-EV-0028: ‘La Resolución AAIP 198/2023, publicada el 18 de octubre de 2023 y firmada por la directora Beatriz…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/296189/20231018. Accessed 2026-05-29.; AR-EV-0141: ‘La Resolución AAIP 126/2024, publicada en el Boletín Oficial el 24 de mayo de 2024 y en vigor desde…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/395000-399999/399750/norma.htm. Accessed 2026-05-29.

¹⁴ AR-EV-0010: ‘Según el Informe Anual 2022 de la AAIP, la autoridad inició 491 expedientes por presuntas…’ Data. https://iapp.org/news/a/la-autoridad-de-proteccion-de-datos-de-argentina-publica-su-informe-anual-2022. Accessed 2026-05-29.

¹⁵ AR-EV-0007: ‘La AAIP publica como guía oficial vigente que los países considerados con legislación adecuada son…’ Report. https://www.argentina.gob.ar/aaip/datospersonales/transferencias-internacionales. Accessed 2026-05-29.

¹⁶ AR-EV-0008: ‘La Law No. 27699, sancionada el 9 de noviembre de 2022 y publicada en el Boletín Oficial el 30 de…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/375000-379999/375738/norma.htm. Accessed 2026-05-29.

¹⁷ AR-EV-0004: ‘El Proyecto de Ley de Protección de Datos Personales presentado por el Poder Ejecutivo Nacional en…’ Policy. https://www.argentina.gob.ar/aaip/datospersonales/proyecto-ley-datos-personales. Accessed 2026-05-29.

¹⁸ AR-EV-0011: ‘El régimen vigente de la Law No. 25326 no contempla obligaciones operativas de notificación reglada de…’ Gap. Accessed 2026-05-29.

¹⁹ AR-EV-0141: ‘La Resolución AAIP 126/2024, publicada en el Boletín Oficial el 24 de mayo de 2024 y en vigor desde…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/395000-399999/399750/norma.htm. Accessed 2026-05-29.

²⁰ AR-EV-0010: ‘Según el Informe Anual 2022 de la AAIP, la autoridad inició 491 expedientes por presuntas…’ Data. https://iapp.org/news/a/la-autoridad-de-proteccion-de-datos-de-argentina-publica-su-informe-anual-2022. Accessed 2026-05-29.

²¹ AR-EV-0145: ‘El 19 de diciembre de 2025 la AAIP inició una investigación de oficio ante una presunta filtración…’ CaseStudy. https://www.argentina.gob.ar/noticias/la-aaip-inicio-una-investigacion-de-oficio-ante-presunta-filtracion-masiva-de-datos. Accessed 2026-05-29.

²² AR-EV-0021: ‘El régimen argentino de transferencias internacionales de datos personales se rige por el artículo…’ Regulation. https://www.argentina.gob.ar/transferencias-internacionales. Accessed 2026-05-29.; AR-EV-0005: ‘La AAIP aprobó por Resolución 47/2018 (publicada 25-07-2018, vigente) las Medidas de Seguridad…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/310000-314999/312662/norma.htm. Accessed 2026-05-29.

²³ AR-EV-0012: ‘La Comunicación A 7724 del BCRA, vigente desde el 6 de septiembre de 2023, establece requisitos…’ Regulation. https://www.grantthornton.com.ar/en/insights/articles/2023/communication-a7724-bcra/. Accessed 2026-05-29.; AR-EV-0023: ‘El BCRA emitió la Comunicación A 8401 (13/02/2026) — más reciente que la A 7724 — disponible en el…’ Regulation. https://www.bcra.gob.ar/archivos/Pdfs/comytexord/A8401.pdf. Accessed 2026-05-29.

²⁴ AR-EV-0014: ‘El Poder Ejecutivo Nacional presentó al Congreso el 30 de junio de 2023 un proyecto de ley para…’ Policy. https://iapp.org/news/a/se-presento-ante-el-congreso-nacional-argentino-un-nuevo-proyecto-de-ley-para-reemplazar-la-actual-ley-de-proteccion-de-datos-personales. Accessed 2026-05-29.; AR-EV-0015: ‘El proyecto de ley 1948-D-2025, ingresado a la Cámara de Diputados en 2025, propone reemplazar la…’ Policy. https://www4.hcdn.gob.ar/dependencias/dsecretaria/Periodo2025/PDF2025/TP2025/1948-D-2025.pdf. Accessed 2026-05-29.

²⁵ AR-EV-0013: ‘ARSAT opera el Centro Nacional de Datos en Benavídez (Buenos Aires), única instalación del país con…’ Report. https://www.arsat.com.ar/datacenter/. Accessed 2026-05-29.

²⁶ AR-EV-0020: ‘El gobierno anunció en octubre de 2024 un plan de privatización parcial de ARSAT (hasta el 49% del…’ Policy. https://www.infobae.com/economia/2024/10/08/el-gobierno-buscara-privatizar-el-49-de-la-estatal-arsat-que-podria-salir-a-la-bolsa-en-2025/. Accessed 2026-05-29.

²⁷ AR-EV-0016: ‘Argentina tiene 13 data centers comerciales identificados por CABASE con capacidad superior a 1 MW…’ Data. https://www.telesemana.com/blog/2026/05/06/argentina-busca-atraer-14-data-centers-y-ampliar-su-infraestructura-digital-para-la-era-de-la-ia/. Accessed 2026-05-29.

²⁸ AR-EV-0018: ‘AWS anunció en marzo de 2022 Local Zones en seis ciudades latinoamericanas, incluida Buenos Aires…’ Report. https://aws.amazon.com/blogs/publicsector/aws-announces-local-zones-latin-america/. Accessed 2026-05-29.; AR-EV-0017: ‘El 10 de octubre de 2025 Sur Energy (Argentina) y OpenAI (Estados Unidos) firmaron una carta de…’ CaseStudy. https://chequeado.com/investigaciones/mega-data-centers-en-la-patagonia-promesas-millonarias-y-alerta-por-la-falta-de-regulacion/. Accessed 2026-05-29.

²⁹ AR-EV-0019: ‘El Súper RIGI (Message 181/2026 al HCDN) propone exigir inversiones mínimas de US$ 1.000 millones…’ Policy. https://www.cronista.com/economia-politica/rigi-y-super-rigi-que-cambia-que-se-amplifica-y-que-se-elimina/. Accessed 2026-05-29.

³⁰ AR-EV-0001: ‘El artículo 12 de la Law No. 25326 prohíbe la transferencia de datos personales de cualquier tipo con…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/60000-64999/64790/texact.htm. Accessed 2026-05-29.

³¹ AR-EV-0025: ‘El Decreto Reglamentario 1558/2001 (anexo I, art. 12) faculta a la Dirección Nacional de Protección…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/decreto-1558-2001-70368. Accessed 2026-05-29.

³² AR-EV-0026: ‘La Disposición DNPDP 60-E/2016, dictada el 16 de noviembre de 2016, aprueba dos conjuntos de…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/265000-269999/267922/norma.htm. Accessed 2026-05-29.

³³ AR-EV-0027: ‘La Resolución AAIP 34/2019, del 22 de febrero de 2019 y firmada por Eduardo Andrés Bertoni…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/202373/20190226. Accessed 2026-05-29.

³⁴ AR-EV-0035: ‘En enero de 2024 la Comisión Europea publicó el primer informe de revisión de las decisiones de…’ Policy. https://www.argentina.gob.ar/noticias/argentina-logro-la-nueva-adecuacion-por-parte-de-la-union-europea-para-el-flujo. Accessed 2026-05-29.

³⁵ AR-EV-0041: ‘No se identificó fuente Tier-1 que documente estadísticas cuantitativas específicas del régimen de…’ Gap. Accessed 2026-05-29.

³⁶ AR-EV-0031: ‘El 13 de noviembre de 2025, los presidentes Donald J. Trump y Javier Milei suscribieron el Joint…’ Policy. https://www.whitehouse.gov/briefings-statements/2025/11/joint-statement-on-framework-for-a-united-states-argentina-agreement-on-reciprocal-trade-and-investment/. Accessed 2026-05-29.; AR-EV-0009: ‘El United States Trade Representative (USTR) publicó el 13 de noviembre de 2025 el Fact Sheet del…’ Policy. https://ustr.gov/about/policy-offices/press-office/fact-sheets/2025/november/fact-sheet-united-states-and-argentina-agree-framework-agreement-reciprocal-trade-and-investment. Accessed 2026-05-29.

³⁷ AR-EV-0032: ‘El 5 de febrero de 2026 en Washington D.C., la República Argentina y los Estados Unidos…’ Policy. https://www.cancilleria.gob.ar/es/destacados/argentina-y-estados-unidos-firmaron-un-acuerdo-sobre-comercio-e-inversiones-reciprocos. Accessed 2026-05-29.

³⁸ AR-EV-0040: ‘Microsoft, en su documentación oficial de cumplimiento publicada en learn.microsoft.com, declara…’ Report. https://learn.microsoft.com/en-us/compliance/regulatory/offering-pdpa-argentina. Accessed 2026-05-29.

³⁹ AR-EV-0042: ‘El Decree 117/2016, suscrito el 12 de enero de 2016, instruye a los ministerios, secretarías y…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/decreto-117-2016-257755. Accessed 2026-05-29.; AR-EV-0043: ‘La Law No. 27275 de Derecho de Acceso a la Información Pública, sancionada el 14 de septiembre de…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/265000-269999/265949/norma.htm. Accessed 2026-05-29.; AR-EV-0045: ‘El portal nacional datos.gob.ar registra 1.235 datasets publicados por 42 organizaciones del sector…’ Data. https://datos.gob.ar/. Accessed 2026-05-29.

⁴⁰ AR-EV-0053: ‘Organizaciones civiles argentinas — LA NACION Data junto con ACIJ, Directorio Legislativo y Poder…’ CaseStudy. https://www.lanacion.com.ar/sociedad/periodismo-de-datos-una-iniciativa-y-un-equipo-para-agregar-valor-e-innovar-nid2312192/. Accessed 2026-05-29.

⁴¹ AR-EV-0055: ‘No se identifica fuente Tier-1 que documente reconocimiento estatal explícito de los datos como…’ Gap. Accessed 2026-05-29.

⁴² AR-EV-0054: ‘No se identifica fuente Tier-1 que documente la existencia de un mandato general operativo de…’ Gap. Accessed 2026-05-29.

⁴³ AR-EV-0047: ‘El impuesto PAIS, que añadía una alícuota del 8% sobre suscripciones a servicios digitales del…’ CaseStudy. https://chequeado.com/el-explicador/fin-del-impuesto-pais-que-impacto-tendra-en-las-compras-de-bienes-y-servicios-en-el-exterior-en-el-turismo-y-en-las-importaciones/. Accessed 2026-05-29.

⁴⁴ AR-EV-0057: ‘No se identifica fuente Tier-1 que documente la implementación por parte de Argentina de mecanismos…’ Gap. Accessed 2026-05-29.

⁴⁵ AR-EV-0048: ‘La Comisión Nacional de Defensa de la Competencia (CNDC) creó un Grupo de Investigación y Trabajo…’ CaseStudy. https://www.argentina.gob.ar/noticias/la-cndc-creo-el-grupo-de-investigacion-y-trabajo-sobre-mercados-digitales. Accessed 2026-05-29.; AR-EV-0049: ‘La CNDC investiga 11 mercados de alta concentración: aluminio, acero, petroquímica, comunicaciones…’ CaseStudy. https://www.casarosada.gob.ar/35904-once-mercados-con-altaconcentraci. Accessed 2026-05-29.

⁴⁶ AR-EV-0058: ‘No se identifica fuente Tier-1 que documente la existencia operativa de data trusts, data…’ Gap. Accessed 2026-05-29.

⁴⁷ AR-EV-0044: ‘El Decree 780/2024, firmado el 30 de agosto de 2024 (BO 02-09-2024), reglamentó modificaciones al…’ Regulation. https://www.boletinoficial.gov.ar/detalleAviso/primera/313139/20240902. Accessed 2026-05-29.

⁴⁸ AR-EV-0059: ‘La Law No. 27437 de Compre Argentino y Desarrollo de Proveedores, promulgada en abril de 2018 y…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/310000-314999/310020/norma.htm. Accessed 2026-05-29.; AR-EV-0063: ‘La Law No. 27506 del Régimen de Promoción de la Economía del Conocimiento (publicada en el Boletín…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/320000-324999/324101/norma.htm. Accessed 2026-05-29.

⁴⁹ AR-EV-0060: ‘La Law No. 19640, sancionada en 1972, estableció un régimen aduanero y fiscal especial para Tierra del…’ Regulation. https://prodyambiente.tierradelfuego.gob.ar/regimen-de-promocion-economica-y-fiscal-ley-19-640-2/. Accessed 2026-05-29.

⁵⁰ AR-EV-0062: ‘Argentina no exhibe acuerdos bilaterales de transferencia tecnológica vigentes con jurisdicciones…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/400000-404999/403230/norma.htm. Accessed 2026-05-29.

⁵¹ AR-EV-0063: ‘La Law No. 27506 del Régimen de Promoción de la Economía del Conocimiento (publicada en el Boletín…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/320000-324999/324101/norma.htm. Accessed 2026-05-29.

⁵² AR-EV-0068: ‘Los Lineamientos Estratégicos de Innovación, Ciencia y Tecnología 2025-2027, aprobados por…’ Policy. https://www.argentina.gob.ar/ciencia/plan-nacional-cti/plan-cti. Accessed 2026-05-29.

⁵³ AR-EV-0064: ‘El Instituto Nacional de Tecnología Industrial (INTI), a través de su área de Micro y…’ CaseStudy. https://www.inti.gob.ar/areas/desarrollo-tecnologico-e-innovacion/areas-de-conocimiento/micro-y-nanotecnologias. Accessed 2026-05-29.

⁵⁴ AR-EV-0061: ‘El Decree 333/2025, publicado en el Boletín Oficial el 20 de mayo de 2025, reduce a 8% el Derecho…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/325601/20250520. Accessed 2026-05-29.

⁵⁵ AR-EV-0066: ‘El Grupo Mirgor adquirió en 2021 la operación de Brightstar en Tierra del Fuego, sumando a IATEC y…’ CaseStudy. https://mirgor.com/mirgor-adquiere-brightstar-tierra-del-fuego/. Accessed 2026-05-29.

⁵⁶ AR-EV-0081: ‘La iniciativa de Software Público de la Subsecretaría de Gobierno Digital (ONTI, dependiente de…’ Policy. https://argentina.gob.ar/jefatura/innovacion-publica/ssetic/onti/software-publico. Accessed 2026-05-29.; AR-EV-0082: ‘La Disposición ONTI 2/2019 aprobó el Código de Buenas Prácticas para el desarrollo de software…’ Policy. https://www.boletinoficial.gob.ar/detalleAviso/primera/206660/20190430. Accessed 2026-05-29.; AR-EV-0083: ‘La Provincia de Santa Fe sancionó la Law No. 12360 el 15 de diciembre de 2004, reglamentada por…’ Regulation. https://www.santafe.gob.ar/index.php/web/content/view/full/3601. Accessed 2026-05-29.

⁵⁷ AR-EV-0076: ‘Huayra GNU/Linux es el sistema operativo libre desarrollado en EDUCAR Sociedad del Estado, descrito…’ CaseStudy. https://huayra.educar.gob.ar/. Accessed 2026-05-29.

⁵⁸ AR-EV-0079: ‘ARSAT desplegó la etapa 1 de la Nube Pública Nacional el 1° de abril de 2021, construida sobre…’ CaseStudy. https://www.canal-ar.com.ar/29335-ARSAT-lanzo-la-Nube-Publica-Nacional-y-ofrece-infraestructura-y-servicios-a-demanda.html. Accessed 2026-05-29.

⁵⁹ AR-EV-0086: ‘La organización oficial ‘argob’ en GitHub aloja 43 repositorios; los más populares corresponden a…’ Data. https://github.com/argob. Accessed 2026-05-29.

⁶⁰ AR-EV-0074: ‘El sistema operativo de escritorio dominante en Argentina al mes de abril de 2026 es Windows con…’ Data. https://gs.statcounter.com/os-market-share/desktop/argentina. Accessed 2026-05-29.

⁶¹ AR-EV-0075: ‘El mercado de sistemas operativos móviles en Argentina al mes de abril de 2026 está dominado por…’ Data. https://gs.statcounter.com/os-market-share/mobile/argentina. Accessed 2026-05-29.

⁶² AR-EV-0084: ‘El mercado latinoamericano de sistemas de gestión de bases de datos está dominado por Oracle…’ Data. https://www.informesdeexpertos.com/informes/mercado-latinoamericano-de-sistemas-de-gestion-de-bases-de-datos. Accessed 2026-05-29.

⁶³ AR-EV-0080: ‘ARSAT desplegó Red Hat OpenShift AI para apoyar operaciones de centro de operaciones de red (NOC)…’ CaseStudy. https://www.redhat.com/en/success-stories/arsat. Accessed 2026-05-29.

⁶⁴ AR-EV-0077: ‘In January 2024 the Milei government suspended access to Conectar Igualdad and Educ.ar —…’ CaseStudy. https://www.ambito.com/politica/el-gobierno-bloqueo-las-plataformas-conectar-igualdad-y-educar-n5931668. Accessed 2026-05-29.; AR-EV-0078: ‘El Decree 963/2024, publicado en el Boletín Oficial el 31 de octubre de 2024, designa a Gastón…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/316281/20241031. Accessed 2026-05-29.

⁶⁵ AR-EV-0098: ‘Mercado Libre lideró el ranking de aplicaciones más descargadas en Argentina en 2024 con 11,7…’ CaseStudy. https://mobiletime.la/noticias/23/01/2025/apps-mas-descargadas-2024/. Accessed 2026-05-29.

⁶⁶ AR-EV-0099: ‘Mercado Pago superó los 50 millones de usuarios activos mensuales en LATAM en 2024 y reportó USD…’ Data. https://investor.mercadolibre.com/. Accessed 2026-05-29.

⁶⁷ AR-EV-0101: ‘El segmento de digital banking argentino está concentrado en tres jugadores domésticos: Ualá (6…’ Data. https://fintechnews.am/fintech-argentina/52228/argentinas-top-3-digital-banks-capture-nearly-90-market-share/. Accessed 2026-05-29.

⁶⁸ AR-EV-0093: ‘Mi Argentina, la plataforma de identidad digital ciudadana, supera los 21 millones de personas…’ CaseStudy. https://www.argentina.gob.ar/noticias/se-renueva-mi-argentina-con-mas-y-mejores-servicios-para-la-ciudadania. Accessed 2026-05-29.

⁶⁹ AR-EV-0102: ‘WhatsApp (Meta) tiene una penetración del 93% entre usuarios mayores de 16 años en Argentina, e…’ Data. https://www.infobae.com/tecno/2025/04/04/radiografia-digital-que-hacen-los-argentinos-en-internet-y-en-que-redes-sociales-pasan-mas-tiempo/. Accessed 2026-05-29.

⁷⁰ AR-EV-0103: ‘Google posee el 94,01% del mercado de búsquedas en Argentina a abril de 2026 (Bing 4,13%, Yahoo…’ Data. https://gs.statcounter.com/search-engine-market-share/all/argentina. Accessed 2026-05-29.

⁷¹ AR-EV-0107: ‘No se identificó fuente Tier-1 que documente cifras consolidadas y auditables de participación de…’ Gap. Accessed 2026-05-29.

⁷² AR-EV-0096: ‘El sector argentino de software y servicios informáticos exportó USD 2.674 millones en 2024, un…’ Data. https://cessi.org.ar/2025/05/21/el-software-argentino-genero-mas-de-6-000-empleos-y-alcanzo-un-record-de-exportaciones/. Accessed 2026-05-29.; AR-EV-0097: ‘Argentina’s Law No. 27506 (Régimen de Promoción de la Economía del Conocimiento), in force since 1…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/ley-27506-324101/actualizacion. Accessed 2026-05-29.

⁷³ AR-EV-0104: ‘Auth0, empresa argentina de autenticación digital fundada en 2013, fue adquirida en 2021 por Okta…’ CaseStudy. https://www.infobae.com/economia/2021/06/23/otra-tech-argentina-se-convirtio-en-unicornio-y-es-el-sexto-local-que-hace-y-quien-es-el-millennial-autodidacta-que-la-fundo/. Accessed 2026-05-29.

⁷⁴ AR-EV-0108: ‘La Disposición 1/2021 de la Dirección Nacional de Ciberseguridad crea el Centro Nacional de…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/241077/20210222. Accessed 2026-05-29.

⁷⁵ AR-EV-0109: ‘Durante 2024, el CERT.ar registró 438 incidentes de ciberseguridad, una cifra 15% superior a los…’ Data. https://www.argentina.gob.ar/jefatura/innovacion-ciencia-y-tecnologia/ciberseguridad/informes-de-la-direccion-nacional-de. Accessed 2026-05-29.

⁷⁶ AR-EV-0112: ‘La Decisión Administrativa 641/2021 aprueba los Requisitos Mínimos de Seguridad de la Información…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/350000-354999/351345/norma.htm. Accessed 2026-05-29.

⁷⁷ AR-EV-0114: ‘Por Decree 941/2025 del Poder Ejecutivo (publicado el 2 de enero de 2026 en el Boletín Oficial) se…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/337032/20260102. Accessed 2026-05-29.

⁷⁸ AR-EV-0109: ‘Durante 2024, el CERT.ar registró 438 incidentes de ciberseguridad, una cifra 15% superior a los…’ Data. https://www.argentina.gob.ar/jefatura/innovacion-ciencia-y-tecnologia/ciberseguridad/informes-de-la-direccion-nacional-de. Accessed 2026-05-29.

⁷⁹ AR-EV-0120: ‘No se identificó fuente Tier-1 que documente investigación criptográfica argentina con publicación…’ Gap. Accessed 2026-05-29.

⁸⁰ AR-EV-0121: ‘No se identificó capacidad documentada de ciberseguridad ofensiva soberana (red team gubernamental…’ Gap. Accessed 2026-05-29.

⁸¹ AR-EV-0116: ‘En abril de 2024 se difundió en Telegram una base con datos de aproximadamente 5,7-6 millones de…’ CaseStudy. https://www.lanacion.com.ar/seguridad/la-venden-a-us-3700-hackearon-la-base-de-datos-nacional-de-licencias-de-conducir-y-muestran-las-de-nid16042024/. Accessed 2026-05-29.

⁸² AR-EV-0117: ‘El 25-26 de diciembre de 2024 los sitios oficiales Mi Argentina y la plataforma de la tarjeta SUBE…’ CaseStudy. https://www.infobae.com/politica/2024/12/26/lo-hicimos-por-diversion-los-hackers-que-atacaron-los-sitios-del-gobierno-dijeron-que-no-tenian-segundo-factor-de-autenticacion/. Accessed 2026-05-29.

⁸³ AR-EV-0119: ‘El RENAPER detectó en 2021 el uso indebido de una clave otorgada a un organismo público (Ministerio…’ CaseStudy. https://www.argentina.gob.ar/noticias/el-renaper-detecto-el-uso-indebido-de-una-clave-otorgada-un-organismo-publico-y-formalizo. Accessed 2026-05-29.

⁸⁴ AR-EV-0122: ‘La Law No. 26388 de Delitos Informáticos, sancionada el 4 de junio de 2008, incorporó al Código Penal…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/ley-26388-141790/texto. Accessed 2026-05-29.; AR-EV-0125: ‘Law No. 25506 (2001) establishes the legal validity of digital and electronic signatures in Argentina…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/70000-74999/70749/norma.htm. Accessed 2026-05-29.; AR-EV-0126: ‘Article 15 of Law No. 27078 — which had designated TIC services as ‘essential and strategic public…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/235000-239999/239771/texact.htm. Accessed 2026-05-29.

⁸⁵ AR-EV-0131: ‘Resolución AAIP 161/2023 created the ‘Programa de Transparencia y Protección de Datos Personales en…’ Policy. https://www.boletinoficial.gob.ar/detalleAviso/primera/293363/20230904. Accessed 2026-05-29.; AR-EV-0134: ‘Argentina lacks a horizontal platform-regulation statute. The intermediary-liability bill…’ CaseStudy. https://rest.hcdn.gob.ar/web/proyectos/290487/adjuntos/104934. Accessed 2026-05-29.; AR-EV-0135: ‘Argentina’s cybersecurity-incident response framework rests on dispositions of the Dirección…’ Regulation. https://www.argentina.gob.ar/jefatura/innovacion-publica/direccion-nacional-ciberseguridad/normativa. Accessed 2026-05-29.

⁸⁶ AR-EV-0004: ‘El Proyecto de Ley de Protección de Datos Personales presentado por el Poder Ejecutivo Nacional en…’ Policy. https://www.argentina.gob.ar/aaip/datospersonales/proyecto-ley-datos-personales. Accessed 2026-05-29.; AR-EV-0178: ‘A Personal Data Protection bill (HCDN 1948-D-2025) introduced in the Argentine Chamber of Deputies…’ Analysis. https://iapp.org/news/a/novedades-legislativas-en-argentina-sobre-protecci-n-de-datos-personales-e-inteligencia-artificial. Accessed 2026-05-29.

⁸⁷ AR-EV-0133: ‘The Ministerio de Desregulación y Transformación del Estado (created July 2024 under Federico…’ Data. https://www.argentina.gob.ar/desregulacion. Accessed 2026-05-29.

⁸⁸ AR-EV-0126: ‘Article 15 of Law No. 27078 — which had designated TIC services as ‘essential and strategic public…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/235000-239999/239771/texact.htm. Accessed 2026-05-29.

⁸⁹ AR-EV-0139: ‘The Argentina national AI ecosystem framework — the Plan Nacional de Inteligencia Artificial…’ Report. https://oecd-opsi.org/wp-content/uploads/2021/02/Argentina-National-AI-Strategy.pdf. Accessed 2026-05-29.

⁹⁰ AR-EV-0132: ‘President Javier Milei has publicly positioned Argentina to compete as a ‘low-regulation AI hub’…’ CaseStudy. https://chequeado.com/el-explicador/argentina-polo-de-inteligencia-artificial-que-propone-el-gobierno-de-javier-milei-y-que-chances-hay-de-que-suceda/. Accessed 2026-05-29.

⁹¹ AR-EV-0140: ‘La Resolución AAIP 126/2024, en vigor desde el 1 de junio de 2024, aprueba el régimen sancionatorio…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/308122/20240524. Accessed 2026-05-29.

⁹² AR-EV-0144: ‘La AAIP sancionó a Google Argentina SRL y Google LLC mediante la Resolución 69/2020 por…’ CaseStudy. https://www.argentina.gob.ar/noticias/sancion-google-por-negar-el-derecho-de-acceso. Accessed 2026-05-29.

⁹³ AR-EV-0148: ‘En 2024 la CNDC emitió 89 decisiones en control de concentraciones y una única multa por…’ Data. https://legalblogs.wolterskluwer.com/competition-blog/main-developments-in-competition-law-and-policy-2024-argentina/. Accessed 2026-05-29.

⁹⁴ AR-EV-0142: ‘La Comisión Nacional de Defensa de la Competencia (CNDC) archivó el 2 de julio de 2025 la…’ CaseStudy. https://www.lanacion.com.ar/tecnologia/la-cndc-archivo-la-investigacion-contra-whatsapp-y-meta-por-presunto-abuso-de-posicion-dominante-nid08072025/. Accessed 2026-05-29.

⁹⁵ AR-EV-0158: ‘IRAM (Instituto Argentino de Normalización y Certificación), founded in 1935, is Argentina’s sole…’ Report. https://www.iram.org.ar/en/who-we-are/. Accessed 2026-05-29.

⁹⁶ AR-EV-0162: ‘Verónica Marinelli, coordinator of IRAM’s Subcommittee on Information Security, Cybersecurity and…’ CaseStudy. https://www.iram.org.ar/novedades/novedades-en-seguridad-de-la-informacion/. Accessed 2026-05-29.

⁹⁷ AR-EV-0160: ‘Editors from Universidad Tecnológica Nacional (UTN, Santa Fe Regional Faculty) serve as editor or…’ CaseStudy. https://uit.frsf.utn.edu.ar/. Accessed 2026-05-29.

⁹⁸ AR-EV-0161: ‘The official ITU-T Study Group 20 management roster for the 2025-2028 study period (period 18)…’ Data. https://www.itu.int/net4/ITU-T/lists/mgmt.aspx?Group=20&Period=18. Accessed 2026-05-29.

⁹⁹ AR-EV-0168: ‘No organisation domiciled in Argentina appears among the 329 (as of May 2026) Member organisations…’ Gap. https://www.w3.org/membership/list/. Accessed 2026-05-29.

¹⁰⁰ AR-EV-0163: ‘Argentina was elected President of CITEL’s Steering Committee (COM/CITEL) at the VII CITEL Assembly…’ Report. https://www.argentina.gob.ar/jefatura/innovacion-publica/ssetic/citel. Accessed 2026-05-29.

¹⁰¹ AR-EV-0171: ‘In November 2024 Beatriz Anchorena, head of Argentina’s Access to Public Information Agency (AAIP)…’ CaseStudy. https://www.argentina.gob.ar/noticias/beatriz-anchorena-titular-de-la-aaip-fue-electa-presidenta-del-comite-del-convenio-108-del. Accessed 2026-05-29.

¹⁰² AR-EV-0170: ‘Argentina participated in the second cycle of the UN Open-Ended Working Group (OEWG) on Security of…’ Report. https://www.argentina.gob.ar/noticias/argentina-participo-en-naciones-unidas-de-un-grupo-de-trabajo-sobre-tic-y-seguridad. Accessed 2026-05-29.

¹⁰³ AR-EV-0178: ‘A Personal Data Protection bill (HCDN 1948-D-2025) introduced in the Argentine Chamber of Deputies…’ Analysis. https://iapp.org/news/a/novedades-legislativas-en-argentina-sobre-protecci-n-de-datos-personales-e-inteligencia-artificial. Accessed 2026-05-29.; AR-EV-0169: ‘On 5 February 2026 the Argentine Foreign Ministry announced the signature of the United…’ Policy. https://www.cancilleria.gob.ar/es/actualidad/noticias/argentina-y-estados-unidos-firmaron-un-acuerdo-sobre-comercio-e-inversiones. Accessed 2026-05-29.

¹⁰⁴ AR-EV-0191: ‘En el Ranking Scimago Institutions 2024, CONICET figura como la principal institución gubernamental…’ Report. https://www.scimagoir.com/rankings.php?country=ARG. Accessed 2026-05-29.

¹⁰⁵ AR-EV-0187: ‘El proyecto QUANTEC del Grupo de Circuitos Cuánticos del Centro Atómico Bariloche (CNEA-CONICET)…’ CaseStudy. https://www.argentina.gob.ar/noticias/cientificos-del-centro-atomico-bariloche-buscan-crear-un-procesador-cuantico-con-circuitos. Accessed 2026-05-29.

¹⁰⁶ AR-EV-0186: ‘Clementina XXI, supercomputadora adquirida por Argentina vía Iniciativa Nacional de Supercómputo…’ CaseStudy. https://www.argentina.gob.ar/noticias/argentina-pone-en-funcionamiento-la-supercomputadora-clementina-xxi. Accessed 2026-05-29.

¹⁰⁷ AR-EV-0188: ‘La convocatoria 2023 de Proyectos de Fortalecimiento de Infraestructura Experimental en CyT…’ Policy. https://www.argentina.gob.ar/ciencia/financiamiento/fort-exper-cytcuanticas-2023. Accessed 2026-05-29.

¹⁰⁸ AR-EV-0193: ‘Existen grupos argentinos publicando trabajos de aprendizaje automático y visión por computadora…’ Analysis. https://scholar.google.com/citations?user=ArqlkTUAAAAJ&hl=en. Accessed 2026-05-29.

¹⁰⁹ AR-EV-0187: ‘El proyecto QUANTEC del Grupo de Circuitos Cuánticos del Centro Atómico Bariloche (CNEA-CONICET)…’ CaseStudy. https://www.argentina.gob.ar/noticias/cientificos-del-centro-atomico-bariloche-buscan-crear-un-procesador-cuantico-con-circuitos. Accessed 2026-05-29.

¹¹⁰ AR-EV-0181: ‘Argentina’s R&D investment fell to 0.216% of GDP in 2024, the lowest historical figure on record…’ Data. https://periferia.com.ar/indicios/la-inversion-en-ciencia-y-tecnologia-llego-a-su-mayor-deterioro-historico-en-2024/. Accessed 2026-05-29.

¹¹¹ AR-EV-0182: ‘Law No. 27614 (Ley de Financiamiento del Sistema Nacional de Ciencia, Tecnología e Innovación…’ Regulation. https://www.iade.org.ar/noticias/ley-de-financiamiento-del-sistema-nacional-de-ciencia-tecnologia-e-innovacion-en-argentina. Accessed 2026-05-29.

¹¹² AR-EV-0198: ‘Universidad Tecnológica Nacional (UTN) produces 42.75% of Argentina’s engineering graduates…’ Data. https://frba.utn.edu.ar/dia-de-la-ingenieria-la-utn-forma-mas-del-40-de-los-ingenieros-que-se-graduan-en-el-pais/. Accessed 2026-05-29.

¹¹³ AR-EV-0198: ‘Universidad Tecnológica Nacional (UTN) produces 42.75% of Argentina’s engineering graduates…’ Data. https://frba.utn.edu.ar/dia-de-la-ingenieria-la-utn-forma-mas-del-40-de-los-ingenieros-que-se-graduan-en-el-pais/. Accessed 2026-05-29.

¹¹⁴ AR-EV-0203: ‘The RAICES Programme (Red de Argentinos Investigadores y Científicos en el Exterior), created in…’ Policy. https://www.argentina.gob.ar/ciencia/seppcti/raices. Accessed 2026-05-29.

¹¹⁵ AR-EV-0207: ‘In the EduRank Latin America Computer Science ranking, UBA is #1 in Argentina and #555 worldwide…’ Data. https://edurank.org/cs/la/. Accessed 2026-05-29.

¹¹⁶ AR-EV-0200: ‘Argentina is in the top 30 nations with highly-skilled emigrants according to OECD data, with…’ Data. https://www.untref.edu.ar/mundountref/argentina-se-convirtio-en-un-polo-de-emigracion. Accessed 2026-05-29.; AR-EV-0202: ‘Argentine remote workers for foreign employers earn between USD 2,500 and USD 5,000 per month on…’ Data. https://www.infobae.com/economia/2024/09/15/los-argentinos-que-trabajan-para-el-exterior-ganan-entre-2500-y-5000-dolares-por-mes/. Accessed 2026-05-29.

¹¹⁷ AR-EV-0197: ‘Argentina’s 2025 budget projects Science and Technology investment at approximately 0.2% of GDP —…’ Data. https://periferia.com.ar/indicios/el-presupuesto-2025-deja-la-inversion-en-ciencia-y-tecnologia-al-nivel-del-2002/. Accessed 2026-05-29.; AR-EV-0196: ‘In 2024 CONICET cut doctoral scholarship allocations by approximately 30%, awarded 950 scholarships…’ Policy. https://www.scidev.net/america-latina/news/mas-becas-en-un-clima-incierto-para-hacer-ciencia-en-argentina/. Accessed 2026-05-29.

¹¹⁸ AR-EV-0212: ‘Argentina’s software industry reached USD 22,221 million in revenue in 2024 (+13.1% interannual)…’ Data. https://www.itsitio.com/ar/software/software-argentino-bate-records/. Accessed 2026-05-29.

¹¹⁹ AR-EV-0222: ‘Argentina’s software industry employs more than 140,000 people formally — exceeding employment in…’ Report. https://fund.ar/publicacion/introduccion-a-la-industria-del-software/. Accessed 2026-05-29.

¹²⁰ AR-EV-0096: ‘El sector argentino de software y servicios informáticos exportó USD 2.674 millones en 2024, un…’ Data. https://cessi.org.ar/2025/05/21/el-software-argentino-genero-mas-de-6-000-empleos-y-alcanzo-un-record-de-exportaciones/. Accessed 2026-05-29.; AR-EV-0214: ‘In 2022 Argentine SBC exports were USD 8,047 million, with software making up 33%, professional…’ Data. https://argendata.fund.ar/topico/servicios-basados-en-el-conocimiento/. Accessed 2026-05-29.

¹²¹ AR-EV-0101: ‘El segmento de digital banking argentino está concentrado en tres jugadores domésticos: Ualá (6…’ Data. https://fintechnews.am/fintech-argentina/52228/argentinas-top-3-digital-banks-capture-nearly-90-market-share/. Accessed 2026-05-29.

¹²² AR-EV-0218: ‘Argentine startups raised USD 412 million across 62 funding rounds in 2024 (46 seed rounds at USD…’ Data. https://www.lanacion.com.ar/economia/negocios/los-emprendedores-argentinos-levantaron-mas-de-us412-millones-en-2024-nid11062025/. Accessed 2026-05-29.

¹²³ AR-EV-0217: ‘Auth0 — co-founded in 2013 by Argentine engineers Eugenio Pace (ITBA) and Matías Woloski…’ CaseStudy. https://www.roadshow.com.ar/como-nacio-auth0-el-unicornio-argentino-que-se-vendio-en-us-6500-millones/. Accessed 2026-05-29.

¹²⁴ AR-EV-0214: ‘In 2022 Argentine SBC exports were USD 8,047 million, with software making up 33%, professional…’ Data. https://argendata.fund.ar/topico/servicios-basados-en-el-conocimiento/. Accessed 2026-05-29.

¹²⁵ AR-EV-0220: ‘Within Argentina’s R&D ecosystem the software industry accounts for 14% of total business R&D…’ Data. https://www.argentina.gob.ar/noticias/el-potencial-de-id-en-sectores-productivos-de-la-economia-del-conocimiento-software-y-ag-0. Accessed 2026-05-29.

¹²⁶ AR-EV-0217: ‘Auth0 — co-founded in 2013 by Argentine engineers Eugenio Pace (ITBA) and Matías Woloski…’ CaseStudy. https://www.roadshow.com.ar/como-nacio-auth0-el-unicornio-argentino-que-se-vendio-en-us-6500-millones/. Accessed 2026-05-29.

¹²⁷ AR-EV-0215: ‘Argentina has been losing global software-export market share — from 0.6% of global sales in 2011…’ Analysis. https://www.lanacion.com.ar/economia/despertar-el-potencial-de-la-industria-del-software-una-estrategia-para-el-futuro-nid14082024/. Accessed 2026-05-29.

¹²⁸ AR-EV-0218: ‘Argentine startups raised USD 412 million across 62 funding rounds in 2024 (46 seed rounds at USD…’ Data. https://www.lanacion.com.ar/economia/negocios/los-emprendedores-argentinos-levantaron-mas-de-us412-millones-en-2024-nid11062025/. Accessed 2026-05-29.

¹²⁹ AR-EV-0219: ‘Law No. 27742 ‘Bases y Puntos de Partida para la Libertad de los Argentinos’ (sanctioned July 2024)…’ Policy. https://chequeado.com/el-explicador/openai-invertira-us-25-000-millones-en-un-data-center-para-ia-en-la-argentina-las-claves-del-anuncio/. Accessed 2026-05-29.

¹³⁰ AR-EV-0225: ‘La Law No. 27738, sancionada el 10 de octubre de 2023 y publicada en el Boletín Oficial el 23 de…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/296574/20231023. Accessed 2026-05-29.

¹³¹ AR-EV-0230: ‘La Agenda Digital 2030 fue presentada por el gobierno de Mauricio Macri el 5 de noviembre de 2018…’ Policy. https://www.argentina.gob.ar/noticias/el-gobierno-presento-la-nueva-agenda-digital-2030. Accessed 2026-05-29.

¹³² AR-EV-0226: ‘La Resolución 282/2025 de la Secretaría de Innovación, Ciencia y Tecnología (Jefatura de Gabinete)…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/333751/20251031. Accessed 2026-05-29.

¹³³ AR-EV-0227: ‘La Decisión Administrativa 899/2024, publicada el 24 de septiembre de 2024, crea la Mesa…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/314465/20240924. Accessed 2026-05-29.

¹³⁴ AR-EV-0227: ‘La Decisión Administrativa 899/2024, publicada el 24 de septiembre de 2024, crea la Mesa…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/314465/20240924. Accessed 2026-05-29.

¹³⁵ AR-EV-0229: ‘El Banco Interamericano de Desarrollo aprobó en 2023 un préstamo por USD 35 millones a Argentina…’ CaseStudy. https://www.argentina.gob.ar/noticias/nuevo-programa-de-35-millones-de-dolares-para-el-desarrollo-de-la-inteligencia-artificial. Accessed 2026-05-29.

¹³⁶ AR-EV-0235: ‘La Función Ciencia y Tecnología del Presupuesto Nacional cayó un 11,4% adicional en el primer…’ Data. https://ciicti.org/el-2026-llego-con-mas-recortes-presupuestarios-a-la-ciencia-argentina/. Accessed 2026-05-29.

¹³⁷ AR-EV-0240: ‘El Poder Legislativo no sancionó la Ley de Presupuesto Nacional 2025: el Decree 425/2025 y la…’ Data. https://chequeado.com/el-explicador/presupuesto-2025-las-5-claves-para-entender-la-prorroga-definida-por-el-gobierno-de-javier-milei/. Accessed 2026-05-29.

¹³⁸ AR-EV-0236: ‘El programa Conectar Igualdad —creado en 2010 por Decree 459/10, suspendido durante la…’ CaseStudy. https://palabrasdelderecho.com.ar/articulo/5007/Prorrogaron-la-vigencia-del-Programa-Conectar-Igualdad-por-dos-meses. Accessed 2026-05-29.

¹³⁹ AR-EV-0234: ‘El Ministerio de Ciencia, Tecnología e Innovación fue creado por Decree 21/2007 bajo la…’ CaseStudy. https://es.wikipedia.org/wiki/Ministerio_de_Ciencia,Tecnolog%C3%ADa_e_Innovaci%C3%B3n(Argentina). Accessed 2026-05-29.

¹⁴⁰ AR-EV-0232: ‘El Decreto de Necesidad y Urgencia 70/2023, emitido el 20 de diciembre de 2023, declaró la…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/301122/20231221. Accessed 2026-05-29.

¹⁴¹ AR-EV-0237: ‘El gobierno argentino articula públicamente la ambición de convertir al país en ‘uno de los cuatro…’ Policy. https://www.telesemana.com/blog/2024/12/02/argentina-trabaja-en-desregulacion-energia-y-analisis-de-marcos-normativos-para-convertirse-en-el-hub-de-la-inteligencia-artificial-de-sudamerica/. Accessed 2026-05-29.

¹⁴² AR-EV-0238: ‘La fundación Fundar argumenta que la estrategia de IA del gobierno Milei tiene un ‘talón de…’ Analysis. https://fund.ar/publicacion/la-estrategia-de-milei-en-inteligencia-artificial-tiene-un-talon-de-aquiles/. Accessed 2026-05-29.

¹⁴³ AR-EV-0031: ‘El 13 de noviembre de 2025, los presidentes Donald J. Trump y Javier Milei suscribieron el Joint…’ Policy. https://www.whitehouse.gov/briefings-statements/2025/11/joint-statement-on-framework-for-a-united-states-argentina-agreement-on-reciprocal-trade-and-investment/. Accessed 2026-05-29.; AR-EV-0032: ‘El 5 de febrero de 2026 en Washington D.C., la República Argentina y los Estados Unidos…’ Policy. https://www.cancilleria.gob.ar/es/destacados/argentina-y-estados-unidos-firmaron-un-acuerdo-sobre-comercio-e-inversiones-reciprocos. Accessed 2026-05-29.

¹⁴⁴ AR-EV-0133: ‘The Ministerio de Desregulación y Transformación del Estado (created July 2024 under Federico…’ Data. https://www.argentina.gob.ar/desregulacion. Accessed 2026-05-29.

¹⁴⁵ AR-EV-0235: ‘La Función Ciencia y Tecnología del Presupuesto Nacional cayó un 11,4% adicional en el primer…’ Data. https://ciicti.org/el-2026-llego-con-mas-recortes-presupuestarios-a-la-ciencia-argentina/. Accessed 2026-05-29.; AR-EV-0182: ‘Law No. 27614 (Ley de Financiamiento del Sistema Nacional de Ciencia, Tecnología e Innovación…’ Regulation. https://www.iade.org.ar/noticias/ley-de-financiamiento-del-sistema-nacional-de-ciencia-tecnologia-e-innovacion-en-argentina. Accessed 2026-05-29.

¹⁴⁶ AR-EV-0035: ‘En enero de 2024 la Comisión Europea publicó el primer informe de revisión de las decisiones de…’ Policy. https://www.argentina.gob.ar/noticias/argentina-logro-la-nueva-adecuacion-por-parte-de-la-union-europea-para-el-flujo. Accessed 2026-05-29.

¹⁴⁷ AR-EV-0007: ‘La AAIP publica como guía oficial vigente que los países considerados con legislación adecuada son…’ Report. https://www.argentina.gob.ar/aaip/datospersonales/transferencias-internacionales. Accessed 2026-05-29.

¹⁴⁸ AR-EV-0018: ‘AWS anunció en marzo de 2022 Local Zones en seis ciudades latinoamericanas, incluida Buenos Aires…’ Report. https://aws.amazon.com/blogs/publicsector/aws-announces-local-zones-latin-america/. Accessed 2026-05-29.; AR-EV-0219: ‘Law No. 27742 ‘Bases y Puntos de Partida para la Libertad de los Argentinos’ (sanctioned July 2024)…’ Policy. https://chequeado.com/el-explicador/openai-invertira-us-25-000-millones-en-un-data-center-para-ia-en-la-argentina-las-claves-del-anuncio/. Accessed 2026-05-29.

¹⁴⁹ AR-EV-0104: ‘Auth0, empresa argentina de autenticación digital fundada en 2013, fue adquirida en 2021 por Okta…’ CaseStudy. https://www.infobae.com/economia/2021/06/23/otra-tech-argentina-se-convirtio-en-unicornio-y-es-el-sexto-local-que-hace-y-quien-es-el-millennial-autodidacta-que-la-fundo/. Accessed 2026-05-29.; AR-EV-0217: ‘Auth0 — co-founded in 2013 by Argentine engineers Eugenio Pace (ITBA) and Matías Woloski…’ CaseStudy. https://www.roadshow.com.ar/como-nacio-auth0-el-unicornio-argentino-que-se-vendio-en-us-6500-millones/. Accessed 2026-05-29.

¹⁵⁰ AR-EV-0200: ‘Argentina is in the top 30 nations with highly-skilled emigrants according to OECD data, with…’ Data. https://www.untref.edu.ar/mundountref/argentina-se-convirtio-en-un-polo-de-emigracion. Accessed 2026-05-29.

¹⁵¹ AR-EV-0181: ‘Argentina’s R&D investment fell to 0.216% of GDP in 2024, the lowest historical figure on record…’ Data. https://periferia.com.ar/indicios/la-inversion-en-ciencia-y-tecnologia-llego-a-su-mayor-deterioro-historico-en-2024/. Accessed 2026-05-29.

¹⁵² AR-EV-0212: ‘Argentina’s software industry reached USD 22,221 million in revenue in 2024 (+13.1% interannual)…’ Data. https://www.itsitio.com/ar/software/software-argentino-bate-records/. Accessed 2026-05-29.; AR-EV-0001: ‘El artículo 12 de la Law No. 25326 prohíbe la transferencia de datos personales de cualquier tipo con…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/60000-64999/64790/texact.htm. Accessed 2026-05-29.

¹⁵³ AR-EV-0031: ‘El 13 de noviembre de 2025, los presidentes Donald J. Trump y Javier Milei suscribieron el Joint…’ Policy. https://www.whitehouse.gov/briefings-statements/2025/11/joint-statement-on-framework-for-a-united-states-argentina-agreement-on-reciprocal-trade-and-investment/. Accessed 2026-05-29.; AR-EV-0055: ‘No se identifica fuente Tier-1 que documente reconocimiento estatal explícito de los datos como…’ Gap. Accessed 2026-05-29.

¹⁵⁴ AR-EV-0064: ‘El Instituto Nacional de Tecnología Industrial (INTI), a través de su área de Micro y…’ CaseStudy. https://www.inti.gob.ar/areas/desarrollo-tecnologico-e-innovacion/areas-de-conocimiento/micro-y-nanotecnologias. Accessed 2026-05-29.; AR-EV-0080: ‘ARSAT desplegó Red Hat OpenShift AI para apoyar operaciones de centro de operaciones de red (NOC)…’ CaseStudy. https://www.redhat.com/en/success-stories/arsat. Accessed 2026-05-29.; AR-EV-0200: ‘Argentina is in the top 30 nations with highly-skilled emigrants according to OECD data, with…’ Data. https://www.untref.edu.ar/mundountref/argentina-se-convirtio-en-un-polo-de-emigracion. Accessed 2026-05-29.

¹⁵⁵ AR-EV-0133: ‘The Ministerio de Desregulación y Transformación del Estado (created July 2024 under Federico…’ Data. https://www.argentina.gob.ar/desregulacion. Accessed 2026-05-29.; AR-EV-0235: ‘La Función Ciencia y Tecnología del Presupuesto Nacional cayó un 11,4% adicional en el primer…’ Data. https://ciicti.org/el-2026-llego-con-mas-recortes-presupuestarios-a-la-ciencia-argentina/. Accessed 2026-05-29.

¹⁵⁶ AR-EV-0035: ‘En enero de 2024 la Comisión Europea publicó el primer informe de revisión de las decisiones de…’ Policy. https://www.argentina.gob.ar/noticias/argentina-logro-la-nueva-adecuacion-por-parte-de-la-union-europea-para-el-flujo. Accessed 2026-05-29.

¹⁵⁷ AR-EV-0004: ‘El Proyecto de Ley de Protección de Datos Personales presentado por el Poder Ejecutivo Nacional en…’ Policy. https://www.argentina.gob.ar/aaip/datospersonales/proyecto-ley-datos-personales. Accessed 2026-05-29.; AR-EV-0141: ‘La Resolución AAIP 126/2024, publicada en el Boletín Oficial el 24 de mayo de 2024 y en vigor desde…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/395000-399999/399750/norma.htm. Accessed 2026-05-29.

¹⁵⁸ AR-EV-0032: ‘El 5 de febrero de 2026 en Washington D.C., la República Argentina y los Estados Unidos…’ Policy. https://www.cancilleria.gob.ar/es/destacados/argentina-y-estados-unidos-firmaron-un-acuerdo-sobre-comercio-e-inversiones-reciprocos. Accessed 2026-05-29.

¹⁵⁹ AR-EV-0227: ‘La Decisión Administrativa 899/2024, publicada el 24 de septiembre de 2024, crea la Mesa…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/314465/20240924. Accessed 2026-05-29.; AR-EV-0182: ‘Law No. 27614 (Ley de Financiamiento del Sistema Nacional de Ciencia, Tecnología e Innovación…’ Regulation. https://www.iade.org.ar/noticias/ley-de-financiamiento-del-sistema-nacional-de-ciencia-tecnologia-e-innovacion-en-argentina. Accessed 2026-05-29.; AR-EV-0240: ‘El Poder Legislativo no sancionó la Ley de Presupuesto Nacional 2025: el Decree 425/2025 y la…’ Data. https://chequeado.com/el-explicador/presupuesto-2025-las-5-claves-para-entender-la-prorroga-definida-por-el-gobierno-de-javier-milei/. Accessed 2026-05-29.

¹⁶⁰ AR-EV-0097: ‘Argentina’s Law No. 27506 (Régimen de Promoción de la Economía del Conocimiento), in force since 1…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/ley-27506-324101/actualizacion. Accessed 2026-05-29.; AR-EV-0217: ‘Auth0 — co-founded in 2013 by Argentine engineers Eugenio Pace (ITBA) and Matías Woloski…’ CaseStudy. https://www.roadshow.com.ar/como-nacio-auth0-el-unicornio-argentino-que-se-vendio-en-us-6500-millones/. Accessed 2026-05-29.

¹⁶¹ AR-EV-0196: ‘In 2024 CONICET cut doctoral scholarship allocations by approximately 30%, awarded 950 scholarships…’ Policy. https://www.scidev.net/america-latina/news/mas-becas-en-un-clima-incierto-para-hacer-ciencia-en-argentina/. Accessed 2026-05-29.; AR-EV-0186: ‘Clementina XXI, supercomputadora adquirida por Argentina vía Iniciativa Nacional de Supercómputo…’ CaseStudy. https://www.argentina.gob.ar/noticias/argentina-pone-en-funcionamiento-la-supercomputadora-clementina-xxi. Accessed 2026-05-29.; AR-EV-0187: ‘El proyecto QUANTEC del Grupo de Circuitos Cuánticos del Centro Atómico Bariloche (CNEA-CONICET)…’ CaseStudy. https://www.argentina.gob.ar/noticias/cientificos-del-centro-atomico-bariloche-buscan-crear-un-procesador-cuantico-con-circuitos. Accessed 2026-05-29.

¹⁶² AR-EV-0001: ‘El artículo 12 de la Law No. 25326 prohíbe la transferencia de datos personales de cualquier tipo con…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/60000-64999/64790/texact.htm. Accessed 2026-05-29.

¹⁶³ AR-EV-0002: ‘El Decree 1558/2001, promulgado el 29 de noviembre de 2001 y publicado en el Boletín Oficial el 3…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/verNorma.do?id=70368. Accessed 2026-05-29.

¹⁶⁴ AR-EV-0003: ‘La AAIP se autodefine en su sitio oficial como garante de la protección de datos personales y la…’ Regulation. https://www.argentina.gob.ar/aaip/datospersonales. Accessed 2026-05-29.

¹⁶⁵ AR-EV-0004: ‘El Proyecto de Ley de Protección de Datos Personales presentado por el Poder Ejecutivo Nacional en…’ Policy. https://www.argentina.gob.ar/aaip/datospersonales/proyecto-ley-datos-personales. Accessed 2026-05-29.

¹⁶⁶ AR-EV-0005: ‘La AAIP aprobó por Resolución 47/2018 (publicada 25-07-2018, vigente) las Medidas de Seguridad…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/310000-314999/312662/norma.htm. Accessed 2026-05-29.

¹⁶⁷ AR-EV-0007: ‘La AAIP publica como guía oficial vigente que los países considerados con legislación adecuada son…’ Report. https://www.argentina.gob.ar/aaip/datospersonales/transferencias-internacionales. Accessed 2026-05-29.

¹⁶⁸ AR-EV-0008: ‘La Law No. 27699, sancionada el 9 de noviembre de 2022 y publicada en el Boletín Oficial el 30 de…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/375000-379999/375738/norma.htm. Accessed 2026-05-29.

¹⁶⁹ AR-EV-0009: ‘El United States Trade Representative (USTR) publicó el 13 de noviembre de 2025 el Fact Sheet del…’ Policy. https://ustr.gov/about/policy-offices/press-office/fact-sheets/2025/november/fact-sheet-united-states-and-argentina-agree-framework-agreement-reciprocal-trade-and-investment. Accessed 2026-05-29.

¹⁷⁰ AR-EV-0010: ‘Según el Informe Anual 2022 de la AAIP, la autoridad inició 491 expedientes por presuntas…’ Data. https://iapp.org/news/a/la-autoridad-de-proteccion-de-datos-de-argentina-publica-su-informe-anual-2022. Accessed 2026-05-29.

¹⁷¹ AR-EV-0011: ‘El régimen vigente de la Law No. 25326 no contempla obligaciones operativas de notificación reglada de…’ Gap. Accessed 2026-05-29.

¹⁷² AR-EV-0012: ‘La Comunicación A 7724 del BCRA, vigente desde el 6 de septiembre de 2023, establece requisitos…’ Regulation. https://www.grantthornton.com.ar/en/insights/articles/2023/communication-a7724-bcra/. Accessed 2026-05-29.

¹⁷³ AR-EV-0013: ‘ARSAT opera el Centro Nacional de Datos en Benavídez (Buenos Aires), única instalación del país con…’ Report. https://www.arsat.com.ar/datacenter/. Accessed 2026-05-29.

¹⁷⁴ AR-EV-0014: ‘El Poder Ejecutivo Nacional presentó al Congreso el 30 de junio de 2023 un proyecto de ley para…’ Policy. https://iapp.org/news/a/se-presento-ante-el-congreso-nacional-argentino-un-nuevo-proyecto-de-ley-para-reemplazar-la-actual-ley-de-proteccion-de-datos-personales. Accessed 2026-05-29.

¹⁷⁵ AR-EV-0015: ‘El proyecto de ley 1948-D-2025, ingresado a la Cámara de Diputados en 2025, propone reemplazar la…’ Policy. https://www4.hcdn.gob.ar/dependencias/dsecretaria/Periodo2025/PDF2025/TP2025/1948-D-2025.pdf. Accessed 2026-05-29.

¹⁷⁶ AR-EV-0016: ‘Argentina tiene 13 data centers comerciales identificados por CABASE con capacidad superior a 1 MW…’ Data. https://www.telesemana.com/blog/2026/05/06/argentina-busca-atraer-14-data-centers-y-ampliar-su-infraestructura-digital-para-la-era-de-la-ia/. Accessed 2026-05-29.

¹⁷⁷ AR-EV-0017: ‘El 10 de octubre de 2025 Sur Energy (Argentina) y OpenAI (Estados Unidos) firmaron una carta de…’ CaseStudy. https://chequeado.com/investigaciones/mega-data-centers-en-la-patagonia-promesas-millonarias-y-alerta-por-la-falta-de-regulacion/. Accessed 2026-05-29.

¹⁷⁸ AR-EV-0018: ‘AWS anunció en marzo de 2022 Local Zones en seis ciudades latinoamericanas, incluida Buenos Aires…’ Report. https://aws.amazon.com/blogs/publicsector/aws-announces-local-zones-latin-america/. Accessed 2026-05-29.

¹⁷⁹ AR-EV-0019: ‘El Súper RIGI (Message 181/2026 al HCDN) propone exigir inversiones mínimas de US$ 1.000 millones…’ Policy. https://www.cronista.com/economia-politica/rigi-y-super-rigi-que-cambia-que-se-amplifica-y-que-se-elimina/. Accessed 2026-05-29.

¹⁸⁰ AR-EV-0020: ‘El gobierno anunció en octubre de 2024 un plan de privatización parcial de ARSAT (hasta el 49% del…’ Policy. https://www.infobae.com/economia/2024/10/08/el-gobierno-buscara-privatizar-el-49-de-la-estatal-arsat-que-podria-salir-a-la-bolsa-en-2025/. Accessed 2026-05-29.

¹⁸¹ AR-EV-0021: ‘El régimen argentino de transferencias internacionales de datos personales se rige por el artículo…’ Regulation. https://www.argentina.gob.ar/transferencias-internacionales. Accessed 2026-05-29.

¹⁸² AR-EV-0023: ‘El BCRA emitió la Comunicación A 8401 (13/02/2026) — más reciente que la A 7724 — disponible en el…’ Regulation. https://www.bcra.gob.ar/archivos/Pdfs/comytexord/A8401.pdf. Accessed 2026-05-29.

¹⁸³ AR-EV-0025: ‘El Decreto Reglamentario 1558/2001 (anexo I, art. 12) faculta a la Dirección Nacional de Protección…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/decreto-1558-2001-70368. Accessed 2026-05-29.

¹⁸⁴ AR-EV-0026: ‘La Disposición DNPDP 60-E/2016, dictada el 16 de noviembre de 2016, aprueba dos conjuntos de…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/265000-269999/267922/norma.htm. Accessed 2026-05-29.

¹⁸⁵ AR-EV-0027: ‘La Resolución AAIP 34/2019, del 22 de febrero de 2019 y firmada por Eduardo Andrés Bertoni…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/202373/20190226. Accessed 2026-05-29.

¹⁸⁶ AR-EV-0028: ‘La Resolución AAIP 198/2023, publicada el 18 de octubre de 2023 y firmada por la directora Beatriz…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/296189/20231018. Accessed 2026-05-29.

¹⁸⁷ AR-EV-0031: ‘El 13 de noviembre de 2025, los presidentes Donald J. Trump y Javier Milei suscribieron el Joint…’ Policy. https://www.whitehouse.gov/briefings-statements/2025/11/joint-statement-on-framework-for-a-united-states-argentina-agreement-on-reciprocal-trade-and-investment/. Accessed 2026-05-29.

¹⁸⁸ AR-EV-0032: ‘El 5 de febrero de 2026 en Washington D.C., la República Argentina y los Estados Unidos…’ Policy. https://www.cancilleria.gob.ar/es/destacados/argentina-y-estados-unidos-firmaron-un-acuerdo-sobre-comercio-e-inversiones-reciprocos. Accessed 2026-05-29.

¹⁸⁹ AR-EV-0035: ‘En enero de 2024 la Comisión Europea publicó el primer informe de revisión de las decisiones de…’ Policy. https://www.argentina.gob.ar/noticias/argentina-logro-la-nueva-adecuacion-por-parte-de-la-union-europea-para-el-flujo. Accessed 2026-05-29.

¹⁹⁰ AR-EV-0040: ‘Microsoft, en su documentación oficial de cumplimiento publicada en learn.microsoft.com, declara…’ Report. https://learn.microsoft.com/en-us/compliance/regulatory/offering-pdpa-argentina. Accessed 2026-05-29.

¹⁹¹ AR-EV-0041: ‘No se identificó fuente Tier-1 que documente estadísticas cuantitativas específicas del régimen de…’ Gap. Accessed 2026-05-29.

¹⁹² AR-EV-0042: ‘El Decree 117/2016, suscrito el 12 de enero de 2016, instruye a los ministerios, secretarías y…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/decreto-117-2016-257755. Accessed 2026-05-29.

¹⁹³ AR-EV-0043: ‘La Law No. 27275 de Derecho de Acceso a la Información Pública, sancionada el 14 de septiembre de…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/265000-269999/265949/norma.htm. Accessed 2026-05-29.

¹⁹⁴ AR-EV-0044: ‘El Decree 780/2024, firmado el 30 de agosto de 2024 (BO 02-09-2024), reglamentó modificaciones al…’ Regulation. https://www.boletinoficial.gov.ar/detalleAviso/primera/313139/20240902. Accessed 2026-05-29.

¹⁹⁵ AR-EV-0045: ‘El portal nacional datos.gob.ar registra 1.235 datasets publicados por 42 organizaciones del sector…’ Data. https://datos.gob.ar/. Accessed 2026-05-29.

¹⁹⁶ AR-EV-0047: ‘El impuesto PAIS, que añadía una alícuota del 8% sobre suscripciones a servicios digitales del…’ CaseStudy. https://chequeado.com/el-explicador/fin-del-impuesto-pais-que-impacto-tendra-en-las-compras-de-bienes-y-servicios-en-el-exterior-en-el-turismo-y-en-las-importaciones/. Accessed 2026-05-29.

¹⁹⁷ AR-EV-0048: ‘La Comisión Nacional de Defensa de la Competencia (CNDC) creó un Grupo de Investigación y Trabajo…’ CaseStudy. https://www.argentina.gob.ar/noticias/la-cndc-creo-el-grupo-de-investigacion-y-trabajo-sobre-mercados-digitales. Accessed 2026-05-29.

¹⁹⁸ AR-EV-0049: ‘La CNDC investiga 11 mercados de alta concentración: aluminio, acero, petroquímica, comunicaciones…’ CaseStudy. https://www.casarosada.gob.ar/35904-once-mercados-con-altaconcentraci. Accessed 2026-05-29.

¹⁹⁹ AR-EV-0053: ‘Organizaciones civiles argentinas — LA NACION Data junto con ACIJ, Directorio Legislativo y Poder…’ CaseStudy. https://www.lanacion.com.ar/sociedad/periodismo-de-datos-una-iniciativa-y-un-equipo-para-agregar-valor-e-innovar-nid2312192/. Accessed 2026-05-29.

²⁰⁰ AR-EV-0054: ‘No se identifica fuente Tier-1 que documente la existencia de un mandato general operativo de…’ Gap. Accessed 2026-05-29.

²⁰¹ AR-EV-0055: ‘No se identifica fuente Tier-1 que documente reconocimiento estatal explícito de los datos como…’ Gap. Accessed 2026-05-29.

²⁰² AR-EV-0057: ‘No se identifica fuente Tier-1 que documente la implementación por parte de Argentina de mecanismos…’ Gap. Accessed 2026-05-29.

²⁰³ AR-EV-0058: ‘No se identifica fuente Tier-1 que documente la existencia operativa de data trusts, data…’ Gap. Accessed 2026-05-29.

²⁰⁴ AR-EV-0059: ‘La Law No. 27437 de Compre Argentino y Desarrollo de Proveedores, promulgada en abril de 2018 y…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/310000-314999/310020/norma.htm. Accessed 2026-05-29.

²⁰⁵ AR-EV-0060: ‘La Law No. 19640, sancionada en 1972, estableció un régimen aduanero y fiscal especial para Tierra del…’ Regulation. https://prodyambiente.tierradelfuego.gob.ar/regimen-de-promocion-economica-y-fiscal-ley-19-640-2/. Accessed 2026-05-29.

²⁰⁶ AR-EV-0061: ‘El Decree 333/2025, publicado en el Boletín Oficial el 20 de mayo de 2025, reduce a 8% el Derecho…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/325601/20250520. Accessed 2026-05-29.

²⁰⁷ AR-EV-0062: ‘Argentina no exhibe acuerdos bilaterales de transferencia tecnológica vigentes con jurisdicciones…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/400000-404999/403230/norma.htm. Accessed 2026-05-29.

²⁰⁸ AR-EV-0063: ‘La Law No. 27506 del Régimen de Promoción de la Economía del Conocimiento (publicada en el Boletín…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/320000-324999/324101/norma.htm. Accessed 2026-05-29.

²⁰⁹ AR-EV-0064: ‘El Instituto Nacional de Tecnología Industrial (INTI), a través de su área de Micro y…’ CaseStudy. https://www.inti.gob.ar/areas/desarrollo-tecnologico-e-innovacion/areas-de-conocimiento/micro-y-nanotecnologias. Accessed 2026-05-29.

²¹⁰ AR-EV-0066: ‘El Grupo Mirgor adquirió en 2021 la operación de Brightstar en Tierra del Fuego, sumando a IATEC y…’ CaseStudy. https://mirgor.com/mirgor-adquiere-brightstar-tierra-del-fuego/. Accessed 2026-05-29.

²¹¹ AR-EV-0068: ‘Los Lineamientos Estratégicos de Innovación, Ciencia y Tecnología 2025-2027, aprobados por…’ Policy. https://www.argentina.gob.ar/ciencia/plan-nacional-cti/plan-cti. Accessed 2026-05-29.

²¹² AR-EV-0074: ‘El sistema operativo de escritorio dominante en Argentina al mes de abril de 2026 es Windows con…’ Data. https://gs.statcounter.com/os-market-share/desktop/argentina. Accessed 2026-05-29.

²¹³ AR-EV-0075: ‘El mercado de sistemas operativos móviles en Argentina al mes de abril de 2026 está dominado por…’ Data. https://gs.statcounter.com/os-market-share/mobile/argentina. Accessed 2026-05-29.

²¹⁴ AR-EV-0076: ‘Huayra GNU/Linux es el sistema operativo libre desarrollado en EDUCAR Sociedad del Estado, descrito…’ CaseStudy. https://huayra.educar.gob.ar/. Accessed 2026-05-29.

²¹⁵ AR-EV-0077: ‘In January 2024 the Milei government suspended access to Conectar Igualdad and Educ.ar —…’ CaseStudy. https://www.ambito.com/politica/el-gobierno-bloqueo-las-plataformas-conectar-igualdad-y-educar-n5931668. Accessed 2026-05-29.

²¹⁶ AR-EV-0078: ‘El Decree 963/2024, publicado en el Boletín Oficial el 31 de octubre de 2024, designa a Gastón…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/316281/20241031. Accessed 2026-05-29.

²¹⁷ AR-EV-0079: ‘ARSAT desplegó la etapa 1 de la Nube Pública Nacional el 1° de abril de 2021, construida sobre…’ CaseStudy. https://www.canal-ar.com.ar/29335-ARSAT-lanzo-la-Nube-Publica-Nacional-y-ofrece-infraestructura-y-servicios-a-demanda.html. Accessed 2026-05-29.

²¹⁸ AR-EV-0080: ‘ARSAT desplegó Red Hat OpenShift AI para apoyar operaciones de centro de operaciones de red (NOC)…’ CaseStudy. https://www.redhat.com/en/success-stories/arsat. Accessed 2026-05-29.

²¹⁹ AR-EV-0081: ‘La iniciativa de Software Público de la Subsecretaría de Gobierno Digital (ONTI, dependiente de…’ Policy. https://argentina.gob.ar/jefatura/innovacion-publica/ssetic/onti/software-publico. Accessed 2026-05-29.

²²⁰ AR-EV-0082: ‘La Disposición ONTI 2/2019 aprobó el Código de Buenas Prácticas para el desarrollo de software…’ Policy. https://www.boletinoficial.gob.ar/detalleAviso/primera/206660/20190430. Accessed 2026-05-29.

²²¹ AR-EV-0083: ‘La Provincia de Santa Fe sancionó la Law No. 12360 el 15 de diciembre de 2004, reglamentada por…’ Regulation. https://www.santafe.gob.ar/index.php/web/content/view/full/3601. Accessed 2026-05-29.

²²² AR-EV-0084: ‘El mercado latinoamericano de sistemas de gestión de bases de datos está dominado por Oracle…’ Data. https://www.informesdeexpertos.com/informes/mercado-latinoamericano-de-sistemas-de-gestion-de-bases-de-datos. Accessed 2026-05-29.

²²³ AR-EV-0086: ‘La organización oficial argob en GitHub aloja 43 repositorios; los más populares corresponden a…’ Data. https://github.com/argob. Accessed 2026-05-29.

²²⁴ AR-EV-0093: ‘Mi Argentina, la plataforma de identidad digital ciudadana, supera los 21 millones de personas…’ CaseStudy. https://www.argentina.gob.ar/noticias/se-renueva-mi-argentina-con-mas-y-mejores-servicios-para-la-ciudadania. Accessed 2026-05-29.

²²⁵ AR-EV-0096: ‘El sector argentino de software y servicios informáticos exportó USD 2.674 millones en 2024, un…’ Data. https://cessi.org.ar/2025/05/21/el-software-argentino-genero-mas-de-6-000-empleos-y-alcanzo-un-record-de-exportaciones/. Accessed 2026-05-29.

²²⁶ AR-EV-0097: ‘Argentina Law No. 27506 (Régimen de Promoción de la Economía del Conocimiento), in force since 1…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/ley-27506-324101/actualizacion. Accessed 2026-05-29.

²²⁷ AR-EV-0098: ‘Mercado Libre lideró el ranking de aplicaciones más descargadas en Argentina en 2024 con 11,7…’ CaseStudy. https://mobiletime.la/noticias/23/01/2025/apps-mas-descargadas-2024/. Accessed 2026-05-29.

²²⁸ AR-EV-0099: ‘Mercado Pago superó los 50 millones de usuarios activos mensuales en LATAM en 2024 y reportó USD…’ Data. https://investor.mercadolibre.com/. Accessed 2026-05-29.

²²⁹ AR-EV-0101: ‘El segmento de digital banking argentino está concentrado en tres jugadores domésticos: Ualá (6…’ Data. https://fintechnews.am/fintech-argentina/52228/argentinas-top-3-digital-banks-capture-nearly-90-market-share/. Accessed 2026-05-29.

²³⁰ AR-EV-0102: ‘WhatsApp (Meta) tiene una penetración del 93% entre usuarios mayores de 16 años en Argentina, e…’ Data. https://www.infobae.com/tecno/2025/04/04/radiografia-digital-que-hacen-los-argentinos-en-internet-y-en-que-redes-sociales-pasan-mas-tiempo/. Accessed 2026-05-29.

²³¹ AR-EV-0103: ‘Google posee el 94,01% del mercado de búsquedas en Argentina a abril de 2026 (Bing 4,13%, Yahoo…’ Data. https://gs.statcounter.com/search-engine-market-share/all/argentina. Accessed 2026-05-29.

²³² AR-EV-0104: ‘Auth0, empresa argentina de autenticación digital fundada en 2013, fue adquirida en 2021 por Okta…’ CaseStudy. https://www.infobae.com/economia/2021/06/23/otra-tech-argentina-se-convirtio-en-unicornio-y-es-el-sexto-local-que-hace-y-quien-es-el-millennial-autodidacta-que-la-fundo/. Accessed 2026-05-29.

²³³ AR-EV-0107: ‘No se identificó fuente Tier-1 que documente cifras consolidadas y auditables de participación de…’ Gap. Accessed 2026-05-29.

²³⁴ AR-EV-0108: ‘La Disposición 1/2021 de la Dirección Nacional de Ciberseguridad crea el Centro Nacional de…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/241077/20210222. Accessed 2026-05-29.

²³⁵ AR-EV-0109: ‘Durante 2024, el CERT.ar registró 438 incidentes de ciberseguridad, una cifra 15% superior a los…’ Data. https://www.argentina.gob.ar/jefatura/innovacion-ciencia-y-tecnologia/ciberseguridad/informes-de-la-direccion-nacional-de. Accessed 2026-05-29.

²³⁶ AR-EV-0112: ‘La Decisión Administrativa 641/2021 aprueba los Requisitos Mínimos de Seguridad de la Información…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/350000-354999/351345/norma.htm. Accessed 2026-05-29.

²³⁷ AR-EV-0114: ‘Por Decree 941/2025 del Poder Ejecutivo (publicado el 2 de enero de 2026 en el Boletín Oficial) se…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/337032/20260102. Accessed 2026-05-29.

²³⁸ AR-EV-0116: ‘En abril de 2024 se difundió en Telegram una base con datos de aproximadamente 5,7-6 millones de…’ CaseStudy. https://www.lanacion.com.ar/seguridad/la-venden-a-us-3700-hackearon-la-base-de-datos-nacional-de-licencias-de-conducir-y-muestran-las-de-nid16042024/. Accessed 2026-05-29.

²³⁹ AR-EV-0117: ‘El 25-26 de diciembre de 2024 los sitios oficiales Mi Argentina y la plataforma de la tarjeta SUBE…’ CaseStudy. https://www.infobae.com/politica/2024/12/26/lo-hicimos-por-diversion-los-hackers-que-atacaron-los-sitios-del-gobierno-dijeron-que-no-tenian-segundo-factor-de-autenticacion/. Accessed 2026-05-29.

²⁴⁰ AR-EV-0119: ‘El RENAPER detectó en 2021 el uso indebido de una clave otorgada a un organismo público (Ministerio…’ CaseStudy. https://www.argentina.gob.ar/noticias/el-renaper-detecto-el-uso-indebido-de-una-clave-otorgada-un-organismo-publico-y-formalizo. Accessed 2026-05-29.

²⁴¹ AR-EV-0120: ‘No se identificó fuente Tier-1 que documente investigación criptográfica argentina con publicación…’ Gap. Accessed 2026-05-29.

²⁴² AR-EV-0121: ‘No se identificó capacidad documentada de ciberseguridad ofensiva soberana (red team gubernamental…’ Gap. Accessed 2026-05-29.

²⁴³ AR-EV-0122: ‘La Law No. 26388 de Delitos Informáticos, sancionada el 4 de junio de 2008, incorporó al Código Penal…’ Regulation. https://www.argentina.gob.ar/normativa/nacional/ley-26388-141790/texto. Accessed 2026-05-29.

²⁴⁴ AR-EV-0125: ‘Law No. 25506 (2001) establishes the legal validity of digital and electronic signatures in Argentina…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/70000-74999/70749/norma.htm. Accessed 2026-05-29.

²⁴⁵ AR-EV-0126: ‘Article 15 of Law No. 27078 — which had designated TIC services as essential and strategic public…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/235000-239999/239771/texact.htm. Accessed 2026-05-29.

²⁴⁶ AR-EV-0131: ‘Resolución AAIP 161/2023 created the Programa de Transparencia y Protección de Datos Personales en…’ Policy. https://www.boletinoficial.gob.ar/detalleAviso/primera/293363/20230904. Accessed 2026-05-29.

²⁴⁷ AR-EV-0132: ‘President Javier Milei has publicly positioned Argentina to compete as a low-regulation AI hub…’ CaseStudy. https://chequeado.com/el-explicador/argentina-polo-de-inteligencia-artificial-que-propone-el-gobierno-de-javier-milei-y-que-chances-hay-de-que-suceda/. Accessed 2026-05-29.

²⁴⁸ AR-EV-0133: ‘The Ministerio de Desregulación y Transformación del Estado (created July 2024 under Federico…’ Data. https://www.argentina.gob.ar/desregulacion. Accessed 2026-05-29.

²⁴⁹ AR-EV-0134: ‘Argentina lacks a horizontal platform-regulation statute. The intermediary-liability bill…’ CaseStudy. https://rest.hcdn.gob.ar/web/proyectos/290487/adjuntos/104934. Accessed 2026-05-29.

²⁵⁰ AR-EV-0135: ‘Argentina cybersecurity-incident response framework rests on dispositions of the Dirección…’ Regulation. https://www.argentina.gob.ar/jefatura/innovacion-publica/direccion-nacional-ciberseguridad/normativa. Accessed 2026-05-29.

²⁵¹ AR-EV-0139: ‘The Argentina national AI ecosystem framework — the Plan Nacional de Inteligencia Artificial…’ Report. https://oecd-opsi.org/wp-content/uploads/2021/02/Argentina-National-AI-Strategy.pdf. Accessed 2026-05-29.

²⁵² AR-EV-0140: ‘La Resolución AAIP 126/2024, en vigor desde el 1 de junio de 2024, aprueba el régimen sancionatorio…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/308122/20240524. Accessed 2026-05-29.

²⁵³ AR-EV-0141: ‘La Resolución AAIP 126/2024, publicada en el Boletín Oficial el 24 de mayo de 2024 y en vigor desde…’ Regulation. https://servicios.infoleg.gob.ar/infolegInternet/anexos/395000-399999/399750/norma.htm. Accessed 2026-05-29.

²⁵⁴ AR-EV-0142: ‘La Comisión Nacional de Defensa de la Competencia (CNDC) archivó el 2 de julio de 2025 la…’ CaseStudy. https://www.lanacion.com.ar/tecnologia/la-cndc-archivo-la-investigacion-contra-whatsapp-y-meta-por-presunto-abuso-de-posicion-dominante-nid08072025/. Accessed 2026-05-29.

²⁵⁵ AR-EV-0144: ‘La AAIP sancionó a Google Argentina SRL y Google LLC mediante la Resolución 69/2020 por…’ CaseStudy. https://www.argentina.gob.ar/noticias/sancion-google-por-negar-el-derecho-de-acceso. Accessed 2026-05-29.

²⁵⁶ AR-EV-0148: ‘En 2024 la CNDC emitió 89 decisiones en control de concentraciones y una única multa por…’ Data. https://legalblogs.wolterskluwer.com/competition-blog/main-developments-in-competition-law-and-policy-2024-argentina/. Accessed 2026-05-29.

²⁵⁷ AR-EV-0158: ‘IRAM (Instituto Argentino de Normalización y Certificación), founded in 1935, is Argentina sole…’ Report. https://www.iram.org.ar/en/who-we-are/. Accessed 2026-05-29.

²⁵⁸ AR-EV-0160: ‘Editors from Universidad Tecnológica Nacional (UTN, Santa Fe Regional Faculty) serve as editor or…’ CaseStudy. https://uit.frsf.utn.edu.ar/. Accessed 2026-05-29.

²⁵⁹ AR-EV-0161: ‘The official ITU-T Study Group 20 management roster for the 2025-2028 study period (period 18)…’ Data. https://www.itu.int/net4/ITU-T/lists/mgmt.aspx?Group=20&Period=18. Accessed 2026-05-29.

²⁶⁰ AR-EV-0162: ‘Verónica Marinelli, coordinator of IRAM Subcommittee on Information Security, Cybersecurity and…’ CaseStudy. https://www.iram.org.ar/novedades/novedades-en-seguridad-de-la-informacion/. Accessed 2026-05-29.

²⁶¹ AR-EV-0163: ‘Argentina was elected President of CITEL Steering Committee (COM/CITEL) at the VII CITEL Assembly…’ Report. https://www.argentina.gob.ar/jefatura/innovacion-publica/ssetic/citel. Accessed 2026-05-29.

²⁶² AR-EV-0168: ‘No organisation domiciled in Argentina appears among the 329 (as of May 2026) Member organisations…’ Gap. https://www.w3.org/membership/list/. Accessed 2026-05-29.

²⁶³ AR-EV-0169: ‘On 5 February 2026 the Argentine Foreign Ministry announced the signature of the United…’ Policy. https://www.cancilleria.gob.ar/es/actualidad/noticias/argentina-y-estados-unidos-firmaron-un-acuerdo-sobre-comercio-e-inversiones. Accessed 2026-05-29.

²⁶⁴ AR-EV-0170: ‘Argentina participated in the second cycle of the UN Open-Ended Working Group (OEWG) on Security of…’ Report. https://www.argentina.gob.ar/noticias/argentina-participo-en-naciones-unidas-de-un-grupo-de-trabajo-sobre-tic-y-seguridad. Accessed 2026-05-29.

²⁶⁵ AR-EV-0171: ‘In November 2024 Beatriz Anchorena, head of Argentina Access to Public Information Agency (AAIP)…’ CaseStudy. https://www.argentina.gob.ar/noticias/beatriz-anchorena-titular-de-la-aaip-fue-electa-presidenta-del-comite-del-convenio-108-del. Accessed 2026-05-29.

²⁶⁶ AR-EV-0178: ‘A Personal Data Protection bill (HCDN 1948-D-2025) introduced in the Argentine Chamber of Deputies…’ Analysis. https://iapp.org/news/a/novedades-legislativas-en-argentina-sobre-protecci-n-de-datos-personales-e-inteligencia-artificial. Accessed 2026-05-29.

²⁶⁷ AR-EV-0181: ‘Argentina R&D investment fell to 0.216% of GDP in 2024, the lowest historical figure on record…’ Data. https://periferia.com.ar/indicios/la-inversion-en-ciencia-y-tecnologia-llego-a-su-mayor-deterioro-historico-en-2024/. Accessed 2026-05-29.

²⁶⁸ AR-EV-0182: ‘Law No. 27614 (Ley de Financiamiento del Sistema Nacional de Ciencia, Tecnología e Innovación…’ Regulation. https://www.iade.org.ar/noticias/ley-de-financiamiento-del-sistema-nacional-de-ciencia-tecnologia-e-innovacion-en-argentina. Accessed 2026-05-29.

²⁶⁹ AR-EV-0186: ‘Clementina XXI, supercomputadora adquirida por Argentina vía Iniciativa Nacional de Supercómputo…’ CaseStudy. https://www.argentina.gob.ar/noticias/argentina-pone-en-funcionamiento-la-supercomputadora-clementina-xxi. Accessed 2026-05-29.

²⁷⁰ AR-EV-0187: ‘El proyecto QUANTEC del Grupo de Circuitos Cuánticos del Centro Atómico Bariloche (CNEA-CONICET)…’ CaseStudy. https://www.argentina.gob.ar/noticias/cientificos-del-centro-atomico-bariloche-buscan-crear-un-procesador-cuantico-con-circuitos. Accessed 2026-05-29.

²⁷¹ AR-EV-0188: ‘La convocatoria 2023 de Proyectos de Fortalecimiento de Infraestructura Experimental en CyT…’ Policy. https://www.argentina.gob.ar/ciencia/financiamiento/fort-exper-cytcuanticas-2023. Accessed 2026-05-29.

²⁷² AR-EV-0191: ‘En el Ranking Scimago Institutions 2024, CONICET figura como la principal institución gubernamental…’ Report. https://www.scimagoir.com/rankings.php?country=ARG. Accessed 2026-05-29.

²⁷³ AR-EV-0193: ‘Existen grupos argentinos publicando trabajos de aprendizaje automático y visión por computadora…’ Analysis. https://scholar.google.com/citations?user=ArqlkTUAAAAJ&hl=en. Accessed 2026-05-29.

²⁷⁴ AR-EV-0196: ‘In 2024 CONICET cut doctoral scholarship allocations by approximately 30%, awarded 950 scholarships…’ Policy. https://www.scidev.net/america-latina/news/mas-becas-en-un-clima-incierto-para-hacer-ciencia-en-argentina/. Accessed 2026-05-29.

²⁷⁵ AR-EV-0197: ‘Argentina 2025 budget projects Science and Technology investment at approximately 0.2% of GDP —…’ Data. https://periferia.com.ar/indicios/el-presupuesto-2025-deja-la-inversion-en-ciencia-y-tecnologia-al-nivel-del-2002/. Accessed 2026-05-29.

²⁷⁶ AR-EV-0198: ‘Universidad Tecnológica Nacional (UTN) produces 42.75% of Argentina engineering graduates…’ Data. https://frba.utn.edu.ar/dia-de-la-ingenieria-la-utn-forma-mas-del-40-de-los-ingenieros-que-se-graduan-en-el-pais/. Accessed 2026-05-29.

²⁷⁷ AR-EV-0200: ‘Argentina is in the top 30 nations with highly-skilled emigrants according to OECD data, with…’ Data. https://www.untref.edu.ar/mundountref/argentina-se-convirtio-en-un-polo-de-emigracion. Accessed 2026-05-29.

²⁷⁸ AR-EV-0202: ‘Argentine remote workers for foreign employers earn between USD 2,500 and USD 5,000 per month on…’ Data. https://www.infobae.com/economia/2024/09/15/los-argentinos-que-trabajan-para-el-exterior-ganan-entre-2500-y-5000-dolares-por-mes/. Accessed 2026-05-29.

²⁷⁹ AR-EV-0203: ‘The RAICES Programme (Red de Argentinos Investigadores y Científicos en el Exterior), created in…’ Policy. https://www.argentina.gob.ar/ciencia/seppcti/raices. Accessed 2026-05-29.

²⁸⁰ AR-EV-0207: ‘In the EduRank Latin America Computer Science ranking, UBA is #1 in Argentina and #555 worldwide…’ Data. https://edurank.org/cs/la/. Accessed 2026-05-29.

²⁸¹ AR-EV-0212: ‘Argentina software industry reached USD 22,221 million in revenue in 2024 (+13.1% interannual)…’ Data. https://www.itsitio.com/ar/software/software-argentino-bate-records/. Accessed 2026-05-29.

²⁸² AR-EV-0214: ‘In 2022 Argentine SBC exports were USD 8,047 million, with software making up 33%, professional…’ Data. https://argendata.fund.ar/topico/servicios-basados-en-el-conocimiento/. Accessed 2026-05-29.

²⁸³ AR-EV-0215: ‘Argentina has been losing global software-export market share — from 0.6% of global sales in 2011…’ Analysis. https://www.lanacion.com.ar/economia/despertar-el-potencial-de-la-industria-del-software-una-estrategia-para-el-futuro-nid14082024/. Accessed 2026-05-29.

²⁸⁴ AR-EV-0217: ‘Auth0 — co-founded in 2013 by Argentine engineers Eugenio Pace (ITBA) and Matías Woloski…’ CaseStudy. https://www.roadshow.com.ar/como-nacio-auth0-el-unicornio-argentino-que-se-vendio-en-us-6500-millones/. Accessed 2026-05-29.

²⁸⁵ AR-EV-0218: ‘Argentine startups raised USD 412 million across 62 funding rounds in 2024 (46 seed rounds at USD…’ Data. https://www.lanacion.com.ar/economia/negocios/los-emprendedores-argentinos-levantaron-mas-de-us412-millones-en-2024-nid11062025/. Accessed 2026-05-29.

²⁸⁶ AR-EV-0219: ‘Law No. 27742 Bases y Puntos de Partida para la Libertad de los Argentinos (sanctioned July 2024)…’ Policy. https://chequeado.com/el-explicador/openai-invertira-us-25-000-millones-en-un-data-center-para-ia-en-la-argentina-las-claves-del-anuncio/. Accessed 2026-05-29.

²⁸⁷ AR-EV-0220: ‘Within Argentina R&D ecosystem the software industry accounts for 14% of total business R&D…’ Data. https://www.argentina.gob.ar/noticias/el-potencial-de-id-en-sectores-productivos-de-la-economia-del-conocimiento-software-y-ag-0. Accessed 2026-05-29.

²⁸⁸ AR-EV-0222: ‘Argentina software industry employs more than 140,000 people formally — exceeding employment in…’ Report. https://fund.ar/publicacion/introduccion-a-la-industria-del-software/. Accessed 2026-05-29.

²⁸⁹ AR-EV-0225: ‘La Law No. 27738, sancionada el 10 de octubre de 2023 y publicada en el Boletín Oficial el 23 de…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/296574/20231023. Accessed 2026-05-29.

²⁹⁰ AR-EV-0226: ‘La Resolución 282/2025 de la Secretaría de Innovación, Ciencia y Tecnología (Jefatura de Gabinete)…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/333751/20251031. Accessed 2026-05-29.

²⁹¹ AR-EV-0227: ‘La Decisión Administrativa 899/2024, publicada el 24 de septiembre de 2024, crea la Mesa…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/314465/20240924. Accessed 2026-05-29.

²⁹² AR-EV-0229: ‘El Banco Interamericano de Desarrollo aprobó en 2023 un préstamo por USD 35 millones a Argentina…’ CaseStudy. https://www.argentina.gob.ar/noticias/nuevo-programa-de-35-millones-de-dolares-para-el-desarrollo-de-la-inteligencia-artificial. Accessed 2026-05-29.

²⁹³ AR-EV-0230: ‘La Agenda Digital 2030 fue presentada por el gobierno de Mauricio Macri el 5 de noviembre de 2018…’ Policy. https://www.argentina.gob.ar/noticias/el-gobierno-presento-la-nueva-agenda-digital-2030. Accessed 2026-05-29.

²⁹⁴ AR-EV-0232: ‘El Decreto de Necesidad y Urgencia 70/2023, emitido el 20 de diciembre de 2023, declaró la…’ Regulation. https://www.boletinoficial.gob.ar/detalleAviso/primera/301122/20231221. Accessed 2026-05-29.

²⁹⁵ AR-EV-0234: ‘El Ministerio de Ciencia, Tecnología e Innovación fue creado por Decree 21/2007 bajo la…’ CaseStudy. https://es.wikipedia.org/wiki/Ministerio_de_Ciencia,_Tecnolog%C3%ADa_e_Innovaci%C3%B3n_(Argentina). Accessed 2026-05-29.

²⁹⁶ AR-EV-0235: ‘La Función Ciencia y Tecnología del Presupuesto Nacional cayó un 11,4% adicional en el primer…’ Data. https://ciicti.org/el-2026-llego-con-mas-recortes-presupuestarios-a-la-ciencia-argentina/. Accessed 2026-05-29.

²⁹⁷ AR-EV-0236: ‘El programa Conectar Igualdad —creado en 2010 por Decree 459/10, suspendido durante la…’ CaseStudy. https://palabrasdelderecho.com.ar/articulo/5007/Prorrogaron-la-vigencia-del-Programa-Conectar-Igualdad-por-dos-meses. Accessed 2026-05-29.

²⁹⁸ AR-EV-0237: ‘El gobierno argentino articula públicamente la ambición de convertir al país en uno de los cuatro…’ Policy. https://www.telesemana.com/blog/2024/12/02/argentina-trabaja-en-desregulacion-energia-y-analisis-de-marcos-normativos-para-convertirse-en-el-hub-de-la-inteligencia-artificial-de-sudamerica/. Accessed 2026-05-29.

²⁹⁹ AR-EV-0238: ‘La fundación Fundar argumenta que la estrategia de IA del gobierno Milei tiene un talón de…’ Analysis. https://fund.ar/publicacion/la-estrategia-de-milei-en-inteligencia-artificial-tiene-un-talon-de-aquiles/. Accessed 2026-05-29.

³⁰⁰ AR-EV-0240: ‘El Poder Legislativo no sancionó la Ley de Presupuesto Nacional 2025: el Decree 425/2025 y la…’ Data. https://chequeado.com/el-explicador/presupuesto-2025-las-5-claves-para-entender-la-prorroga-definida-por-el-gobierno-de-javier-milei/. Accessed 2026-05-29.

Pablo Martínez Pablo Martínez serves as Coordinator of this report. He is a senior full-stack developer and the lead of the SUDOCU project at the National University of General Sarmiento (UNGS).

Germán Pinazo Germán Pinazo is the President and a professor at the National University of General Sarmiento (UNGS), whose research focuses on political economy and economic development.

Raw transcripts are often only the starting point of content organisation. Whether from meetings, interviews, lectures, or internal discussions, speech-to-text output usually contains spoken-language expressions, repetitions, incomplete sentences, and occasional recognition errors, making it unsuitable for direct reading.

Traditionally, organising these materials required substantial manual effort. Today, large language models can assist with tasks such as extracting key points, improving structure, and refining content.

After transcription, there are generally two common output formats: meeting minutes and formal articles.

Meeting minutes are a faithful record. Their job is to capture discussions, viewpoints, decisions, and action items completely and in the meeting’s own order, so that anyone can trace who said what and why a decision was reached. They are typically itemised and meant to be scanned or searched rather than read straight through.

A formal article, by contrast, is a piece of writing in its own right. Rather than recording the discussion, it selects only the most valuable insights and reorganises them by theme into a continuous, self-contained argument. It can be read and understood without having attended the meeting, and it favours depth, structure, and narrative flow over sheer completeness.

Therefore, before processing a transcript, it is important to determine the intended output. The goal will directly affect both the prompting strategy and the organisation process.

All example outputs in this guide are generated from the same transcript — a recording of the Hands Off Asia webinar held on 30 April 2025.

Generating Meeting Minutes

Meeting minutes are common in business, project management, and academic settings. Their primary purpose is to help readers quickly understand what happened, what issues were discussed, what viewpoints were raised, and what decisions or follow-up actions were agreed upon.

Important Considerations

In practice, the most common problem when generating meeting minutes is not a lack of fluency but omitted information.

Many users will simply copy a transcript tens of thousands of words long into the AI and type a brief instruction like:

Please summarise this meeting.

For short content, this approach may yield decent results. However, when the meeting content is long, the model may overlook certain parts or even hallucinate due to the excessive context length, adding information that was not present in the original text.

A more recommended method is to process the transcript in stages. First, have the model summarise the content chapter by chapter or by time segment, and then merge these summaries into a complete set of minutes. Compared to compressing the full text in one go, this method is generally more stable and makes it easier to spot any omissions.

For large meetings that span several hours, you can even adopt a “three-step method”: first generate segment summaries, then consolidate all summaries, and finally produce the formal meeting minutes. Although this involves more steps, the result tends to be much more reliable.

Using the Web Version of AI to Generate Meeting Minutes

For most users, the simplest approach remains using the web version of AI directly.

Current mainstream models like ChatGPT, Claude, Gemini, and Qwen all support uploading TXT, DOCX, or PDF files. After uploading the transcript, you can simply ask the model to generate the meeting minutes.

To achieve more consistent results, it is advisable to clearly state in the prompt what the minutes should include, such as the meeting background, discussion topics, decisions made, and action items.

An example prompt is as follows:

Please convert this transcript into professional meeting minutes.

Include:

- Meeting overview
- Key discussion topics
- Decisions made
- Action items
- Open questions

Do not omit important information.

Do not invent content.

Keep the logical order of the original meeting.

Compared to using only the brief instruction Please summarise this meeting, this prompt not only ensures that the model accurately retains the meeting’s core information and logical sequence but also automatically completes the structural reorganisation from a word-for-word record to professional meeting minutes. The output is strictly organised according to a standard minutes framework, including Meeting Overview, Key Discussion Topics, Decisions Made, Action Items, and Open Questions, transforming the originally lengthy and scattered discussion into a document that is easy to read, search, and use for subsequent execution.

The complete output is available in the Bandung Circuit repository.

Beyond this, you can further adjust the prompt according to your specific needs. For example, for policy research, media reports, or institutional archiving, you can request a summary version that emphasises core arguments and main conclusions. For academic exchanges, international forums, or specialised seminars, you can retain more of the speakers’ backgrounds, lines of reasoning, and the different participants’ viewpoints to facilitate subsequent research, citation, or content organisation.

When the transcript is particularly long, it is recommended to process it in batches according to the agenda, chapters, or time periods. This not only helps reduce the risk of information loss that can occur when handling very long texts, but also preserves the contextual relationships within each discussion section more accurately, making it easier to verify, supplement, and integrate later, ultimately producing a more complete and accurate set of meeting minutes or event records.

Using a Skill to Generate Meeting Minutes

For users already employing skills within an Agent or Visual Studio Code, the task of generating meeting minutes can be further automated. One commonly used option is the meeting-minutes skill.

This type of skill is usually optimised specifically for meeting scenarios. Besides summarising the discussion, it can also automatically extract decisions, action items, and follow-up tasks.

Installing this skill in an Agent is straightforward. Open the Agent and enter:

Please help me install this skill:

npx skills add https://github.com/github/awesome-copilot --skill meeting-minutes

After the installation is complete, restart Visual Studio Code to refresh the skills list, and you can then invoke it directly. For example:

/meeting-minutes Generate professional meeting minutes from @meeting_transcript.srt

For project meetings, research discussions, and planning sessions, the value of such a specialised skill lies not only in more consistent results but in its built-in information extraction rules tailored to meeting contexts. Compared to standard meeting minutes generated via prompts, a skill will often go further to identify attendees, the meeting agenda, decisions, action plans, risk factors, and follow-up items, organising everything according to a fixed template.

As you can see from the output, the content generated by the skill is no longer just a record of the meeting; it resembles a complete meeting management document. For example, in addition to the usual Meeting Overview, Discussion Topics, and Action Items, it also produces sections like Metadata, Attendance, Agenda, Risks / Blockers, Next Meeting, and Attachments, and supplements each task with an owner, acceptance criteria, and related resource links. This format is much better suited for team collaboration, project tracking, and knowledge archiving, rather than simply documenting what was discussed.

The complete output is available in the Bandung Circuit repository.

Turning Transcripts into Formal Articles

Turning a transcript into an article is not simply a matter of shortening it. A discussion unfolds in the order people happen to speak: arguments are scattered across different voices, points are repeated, and ideas often arrive half-formed before being picked up again later. Writing it up as an article means reshaping that raw material — drawing out the core arguments, grouping related points by theme, cutting repetition and filler, and rebuilding them into a single line of reasoning a reader can follow from beginning to end. The harder part is doing all this while staying faithful to what the speakers actually meant: their terminology, their emphasis, and their stance.

Prompting the AI to Write the Article

1. Defining the Editor Identity

The first step in generating an article is not telling the AI to write, but telling it what identity it should write from. Without a role definition, models tend to default to a generic writing mode: lightly rephrasing sentences, compressing repetitive content, and mechanically organising paragraphs. Actual editing, however, reconstructs spoken material into a complete written argument. You therefore need to explicitly tell the model that it is not a summarisation assistant or a meeting minutes tool, but a professional editor.

## Your Role


You are an experienced editor specialising in transforming raw transcripts into publication-ready articles.

Your task is not to summarise the transcript or produce meeting notes. Instead, you must restructure spoken content into a coherent written argument suitable for public readers while remaining faithful to the original speakers' meaning, terminology, and intent.

The final article should read as a polished piece of analytical writing rather than an edited transcript.

2. Identifying the Content Type

There is no single way to rewrite a transcript. A multi‑speaker panel discussion and a solo lecture have fundamentally different content structures. If the content type is not identified first, the AI can easily apply the wrong organisational approach. Before formal writing begins, therefore, the model should complete a content classification.

## Identify the Transcript Genre


Before writing, determine the primary format of the transcript.

Possible genres include:

- Panel Discussion
- Webinar
- Lecture
- Keynote Speech
- Interview
- Conference Session
- Debate
- Roundtable Discussion

Select the dominant genre and choose an appropriate restructuring strategy.

For example:

- Multi-speaker discussions should be reorganised around themes.
- Lectures should preserve the original argumentative sequence.
- Interviews may remain in Q&A form or be converted into a feature article.
- Debates should preserve disagreements and opposing positions.

The purpose of this step is to make the AI decide on the article’s organising logic before it begins writing.

3. Outlining First Instead of Writing Directly

Many people ask the AI to output a complete article right away. In practice, the most important step in editing is often not writing but planning the structure. A good outline determines the final quality of the article. It is therefore advisable to split the generation process into two phases: plan first, then write.

## Outline First

Before drafting the article, generate a detailed outline.

The outline should include:

- Proposed article title
- Main thesis
- Section headings
- Key arguments under each section
- Possible opening angles

Also identify any information that may require verification, including:

- Personal names
- Organisations
- Locations
- Dates
- Statistics
- Quotations

Do not guess uncertain information.

This approach lets you check whether the structure makes sense before moving into the formal writing stage.

4. Establishing Fidelity Rules

The greatest risk in rewriting transcripts is not language quality but factual errors. When organising content, models tend to automatically supply background knowledge, infer missing information, and even construct logical chains that did not originally exist. Clear source boundaries must therefore be set.

## Source Fidelity Requirements

The article must be based exclusively on the materials provided by the user.

Use only:

- The transcript
- Supporting documents supplied by the user

Do not:

- Search the web
- Add external context
- Invent facts
- Invent quotations
- Introduce unsupported claims

If information is uncertain, flag it instead of guessing.

This section is effectively the most important safety mechanism in the entire skill.

5. Preserving the Original Stance and Expression

Many models exhibit a common tendency: they automatically neutralise positions. Sharp political, social, or academic expressions are rewritten into milder, vaguer language. For news commentary, research institute articles, or political interviews, this often distorts the original meaning. The model therefore needs to be explicitly told to retain the original expression.

## Preserve the Original Voice

Maintain the speakers' terminology, framing, and political language.
Do not soften or neutralise terms used in the source material.

Preserve key concepts exactly as they appear whenever possible.

Avoid introducing artificial balance or alternative viewpoints that do not exist in the source.

The goal of this section is to preserve the author’s thinking, not merely the facts.

6. Specifying the Article Style

Once the content structure and factual boundaries are determined, the final step is style control. This part tells the AI what the finished piece should resemble.

## Writing Style

Write in the style of serious analytical non-fiction.

The article should:

- Read like a research institute publication or long-form commentary.
- Present ideas directly rather than reporting who spoke first.
- Develop a coherent argument.
- Use thematic sections with meaningful headings.
- Maintain a formal and professional tone.
- Conclude with implications, significance, or consequences.

Avoid sounding like:

- Meeting minutes
- Event reports
- Raw transcripts
- Generic AI summaries

Through this layer of constraint, the model shifts from “organising content” to “constructing an article.”

7. Output Format

Finally, the format of the deliverable needs to be clearly specified.

## Output Format

Produce the article as a single Markdown document.

Structure:

# Title

Introduction

## Section One

Content

## Section Two

Content

## Conclusion

Content

Output only the final article in Markdown format.

8. Input Content

Once all the rules above have been defined, the last step is simply to place the transcript in the input area. If you have any supporting materials — background documents, speaker notes, or reference texts — you can include them here too, so the model can draw on them while still staying within the source boundaries set earlier.

## Input Transcript
[Paste the full transcript here.]

## Supporting Materials (optional)
[Paste or attach any background documents, speaker notes, or reference texts here.]

Running this prompt on the transcript produces an article that has completely shed the form of meeting minutes. Content originally presented by different speakers is re‑integrated into a continuous discussion centred on issues such as sovereignty, security, militarisation, and regional peace. This style of writing is closer to academic commentary, thematic analysis, or in‑depth media features:

The complete output is available in the Bandung Circuit repository.

Using a Skill to Generate Articles

Beyond writing prompts directly, a purpose‑built skill can also be used within an Agent workflow to convert transcripts into formal articles. Compared with one‑off prompts, such skills typically come with a more complete rule system, can handle different types of transcript content more consistently, and produce articles with a clear structure, coherent logic, and a style suitable for public reading.

Global South Insights has developed and open-sourced a more fully featured transcript-to-article skill for turning raw transcripts such as meeting notes, interviews, and lecture recordings into high‑quality articles. The skill has built‑in capabilities for article restructuring, logical organisation, language polishing, and style unification, significantly raising the quality of the final output. You can obtain and install this skill from the corresponding repository.

The installation process is very simple. After downloading the skill, just tell the Agent:

Please help me install this skill

The Agent will automatically complete the installation and configuration.

Once installed, you can directly process a transcript file. For example:

/transcript-to-article Create a detailed article outline from @transcript.txt

After execution, the Agent will first analyse the transcript, identify the thematic structure and core arguments, and automatically generate a detailed outline, providing a clear framework for subsequent article writing.

The real difference between this skill and an ordinary prompt is not the quality of the prose but how much of the process it manages for you. A one-off prompt does everything in a single pass and leaves every rule for you to specify. The skill instead runs the job in stages — it produces an outline for your approval before writing any prose — and adds controls a single prompt cannot easily reproduce: it flags the names, dates, and figures it is unsure the transcription got right (drawing on any supporting materials you provide), detects the transcript’s genre and restructures accordingly, and writes to a defined house style you can steer with a sample article, a target length, and a glossary.

The result shows that the article does not simply unfold by country or speaker. Instead, it advances layer by layer around core issues such as “the architecture of militarisation,” “military bases and the security paradox,” “soft power and economic control,” “the price borne by ordinary people,” and “transnational resistance and international solidarity.” Points originally scattered across multiple speakers’ remarks are integrated into a unified chain of argument, making the article closer to a media special report, opinion piece, or an analytical report issued by a research institute.

In addition, this skill automatically generates an HTML presentation page that lays out the article’s structure, key points, and images more clearly, making the piece easier to read, present, and publish online.

The complete output is available in the Bandung Circuit repository.

Turning Transcripts into PowerPoint Presentations

If you need to take meeting minutes, interview records, course content, or seminar transcripts and further use them for sharing and reporting, you might consider organising them into a PowerPoint presentation. Compared to lengthy text, a presentation is more suitable for showcasing core ideas, clarifying the logical structure, and helping the audience quickly grasp the main points.

In our How to Guides series, we have previously covered how to use AI to organise research notes and textual materials into a slide deck. For more details, please refer to How to Present Your Research — From Notes to Slides.

Conclusion

A transcript is essentially raw material, not the finished product. Whether it is a meeting discussion, an academic interview, or a course recording, it always requires further organisation to truly unlock its value.

If your goal is to document the meeting process and preserve the discussion content and decisions, then meeting minutes are usually the more appropriate choice. If your goal is to communicate ideas, write a report, or publish publicly, you need to further distill the content and restructure it according to the readers’ needs.

As the capabilities of large language models continue to improve, a relatively complete workflow has emerged, spanning from audio transcription to meeting minutes and finally to formal article generation. With well-designed prompts and a layer of human review, a single recording can now become usable minutes or a publishable article in a fraction of the time the work once took — cutting the human effort and cost of turning a meeting into something worth reading, and freeing people to focus on the ideas rather than the write-up.

In academic research, meeting minutes, interview collation, and video subtitle production, converting audio or video content into editable text quickly is one of the most common and fundamental data organisation tasks. Compared with manual dictation, using speech recognition tools is not only more efficient, but also more convenient for subsequent proofreading, retrieval, and archiving.

This article will introduce how to transcribe interviews, meetings, and other audio and video materials into text using several common tools, while preserving a relatively complete graphical operation workflow for practical use.

The example recording used in this guide is a guest presentation at the Global South Academic Forum.

Transcribing with Standalone Software

Standalone software is usually the most beginner-friendly approach: you download and install a desktop application, then complete the entire workflow — from upload to export — within a single interface. The tool we recommend here is iFLYREC.

It provides a relatively complete graphical transcription solution, integrating speech recognition, speaker diarisation, domain-specific optimisation, and result export on the same platform. For users who want to avoid script configuration and still require high transcription efficiency, this kind of tool is very suitable for daily office work and research organisation.

Its advantage lies in its clear workflow: users only need to upload a file, select language and scene parameters, and the system will automatically generate a preliminary text. Afterwards, proofreading can be carried out with the help of audio-text synchronised playback, timestamp positioning, and speaker labelling, thereby reducing manual rework. The following uses the transcription of audio and video files as an example to illustrate its basic usage.

1. Software Installation and Account Login

First, visit the official iFLYREC download page:

https://www.iflyrec.com/zhuanwenzi.html

After entering the page, click ‘Download’ to obtain the latest client installation program.

Once the download is complete, double-click the installation package to start the installer and follow the installation wizard to complete the software deployment step by step.

When launching the software for the first time, you need to read and agree to the user agreement, and then log in to the system using your iFLYTEK account.

After logging in, you can enter the software’s main interface and start the subsequent transcription tasks.

2. Overview of the Main Interface

The iFLYREC homepage integrates various speech-to-text capabilities, covering usage needs in different scenarios. The interface provides relatively direct entry points for tasks such as real-time recording, existing file transcription, and subtitle creation, and the usage logic is relatively clear.

The Recent Files area at the bottom of the software displays recently processed tasks, allowing users to review historical files at any time. Once tasks are synchronised to the cloud, it is also convenient to continue processing on different devices, reducing the burden of local file management.

In practical applications, the most common need is not real-time recording, but converting already recorded audio and video materials into text. Therefore, this article focuses on the Import File function.

3. File Upload and Parameter Configuration

Transcription quality is not only closely related to the clarity of the audio itself but also to the initial parameter settings. Before uploading a file, setting options like language, speaker number, and professional domain can usually make the recognition results more stable and simplify subsequent proofreading.

3.1 Selecting the Source Language

The source language setting is an important factor affecting recognition accuracy. Users should choose a language type that matches the audio content as closely as possible, such as English, Chinese-English mixed, or Chinese. When the language is selected incorrectly, the system often produces more errors, especially in cases involving proper nouns and complex long sentence structures.

Clicking More allows you to view a wider range of language options, including Spanish, Japanese, Chinese, and many other languages, as well as enhanced models like English Pro and Chinese-English Mixed Pro.

In actual use, choosing the right language model accurately is often more important than repeatedly making corrections later. For content with many specialised terms, non-standard pronunciation, or mixed-language expressions, selecting an appropriate model can significantly improve readability.

3.2 Importing Audio/Video Files

After completing the language selection, you can drag the file to be processed directly into the upload area, or import it through the file selection window.

It supports a wide range of video and audio formats, covering the vast majority of meeting recordings, course videos, and interview materials. For longer materials, it is recommended to confirm the file naming and source first, so that they are easier to identify during subsequent organisation and export.

3.3 Setting the Number of Speakers

In multi-person interviews or meetings, distinguishing between different speakers is very important for subsequent analysis. The clearer the speaker diarisation, the easier the final text is to read and the more convenient it is for topic summarisation and viewpoint comparison during research.

iFLYREC offers a speaker diarisation feature, allowing users to manually specify the number of speakers or select Auto to let the system determine it automatically.

For example, if the file is a single-person speech, course recording, or podcast interview, you can directly select 1 Speaker. For multi-person discussion scenarios, it is recommended to set the number of people as accurately as possible to reduce errors in the system’s allocation of dialogue turns. More accurate speaker information typically leads to a clearer and more standardised transcription result.

3.4 Configuring Professional Domains

For specialised scenarios such as law, economics, and medicine, iFLYREC also provides domain-specific optimisation models.

When the transcription content involves many specialised terms, selecting the corresponding domain can often effectively reduce the term recognition error rate and improve the overall readability of the text. For instance, if company names, research jargon, or industry abbreviations frequently appear in a meeting, the system is more likely to make reasonable judgments if it knows the general domain.

If the system’s provided domain categories cannot fully match your research topic, you can also use the Keyword Optimization feature to customise keywords for auxiliary optimisation.

This type of setting is particularly suitable for academic interviews, industry exchanges, and thematic discussions. The system will combine keyword information to make more targeted corrections to the recognition results, bringing the final text closer to the original semantics.

3.5 Submitting the Transcription Task

After completing the parameter settings, click Submit to upload the file and submit the task.

During the upload process, the system will display real-time progress. Once the upload is complete, the original Uploading status will change to Open File.

Click it to enter the transcription interface and start processing the task. This process may take some time for large files or long recordings, but the overall operation logic remains relatively intuitive.

4. Viewing Transcription Results and Quality Proofreading

While a task is running or after transcription is complete, the system will display the corresponding file in the left taskbar.

Users can open the task at any time to view the recognition results. iFLYREC uses an audio-text linkage method for proofreading: when you click on any paragraph in the text, the player will automatically jump to the corresponding time position.

This design is very practical for collating interview materials, verifying meeting minutes, and performing sentence-by-sentence proofreading in academic research. It helps users quickly confirm who said a particular sentence, whether the original speech was recognised correctly, and whether the context segmentation is reasonable.

4.1 Preview Settings

The settings menu next to the player offers several auxiliary reading functions.

Among them, Display Speaker shows identity tags for different speakers; Speaker Filtering can filter playback content by speaker; Display Timecode shows the timestamps corresponding to text paragraphs; and Skip Silent Segments can automatically skip silent parts, making reading and playback more compact.

These settings are particularly helpful when dealing with long meetings, in-depth interviews, and classroom recordings. For users who need to quickly locate content, they not only improve efficiency but also reduce the hassle of repeatedly dragging the progress bar.

4.2 Automatic Translation Function

In addition to speech-to-text, iFLYREC also provides full-text translation capabilities. Clicking the translation button in the interface will perform an automatic translation of the transcribed text.

For cross-language interviews, international meetings, and foreign language course collation, this function can significantly reduce the manual translation workload and makes it convenient to obtain a readable Chinese version first for a second round of revision.

It should be noted that iFLYREC primarily supports translating multiple languages into Chinese. If you need to generate versions in other target languages, it is recommended to process them further with professional translation tools after export to ensure accuracy of expression.

5. Exporting and Saving Results

After completing the proofreading, you can go to the Downloads page to export the final results.

The system supports multiple output formats, including DOCX (Word Document), TXT (Plain Text), and SRT (Subtitle File). You can also decide whether to retain metadata such as timecodes and speaker information as needed.

After confirming the parameters, click Download to save the file to your local device.

For video post-production, the SRT format is usually the most commonly used subtitle output format and can be directly imported into most editing software for subsequent editing.

From a practical application perspective, iFLYREC can well meet the needs of scenarios such as meeting recording, interview collation, lecture transcription, and video subtitle production. Its advantage lies not only in its high recognition efficiency but also in its integration of speech recognition, speaker diarisation, professional term optimisation, translation, and export into a complete workflow, lowering the barrier for non-technical users to carry out speech-to-text tasks.

For researchers, media workers, and content creators who frequently need to process large amounts of audio and video materials, reasonably configuring the language, speaker, and domain parameters, and combining this with a post-processing proofreading workflow, usually allows them to quickly obtain high-quality, structured text results.

Transcribing with Web-Based AI Tools

In addition to standalone transcription software, some AI web interfaces also provide online recording and meeting minutes functions. Qwen’s meeting minutes feature falls into this category. It can record content in real-time during a meeting, categorise speech based on different speakers, and finally generate a relatively complete text summary.

After entering the Qwen official website at https://www.qianwen.com/chat, click ‘More’ → ‘Meeting Minutes’ to enter the recording page.

Once on the meeting minutes page, you need to grant microphone permission so that the system can record audio normally.

During the recording process, the system not only records the speaking time but also distinguishes between different speakers based on timbre and voiceprint characteristics, making it quite practical for multi-person discussion scenarios.

After clicking stop, Qwen will automatically summarise the content and generate a meeting report.

This approach is suitable for impromptu meetings, online discussions, and quick minute-taking, especially for users who don’t want to install extra software. Its advantage lies in its quick learning curve and short workflow, making it ideal for immediate use.

Transcribing with a Local Agent Skill

In addition to cloud-based online tools, you can also complete audio transcription by configuring skills in a local Agent. For example, OpenAI’s Whisper skill is a relatively common type of transcription tool that can run locally and is suitable for users with higher requirements for privacy and workflow.

The method for installing the Whisper skill in an Agent is also quite straightforward. Open the Agent and type:

Please help me install this skill:

npx skills add steipete/clawdis@openai-whisper

After waiting for the installation to complete, restart VS Code to refresh the skills list. Then, when you need to transcribe an audio or video file, simply input:

/openai-whisper Transcribe the audio/video file @file_to_transcribe.mp4 and output the result as a plain text (.txt) transcript.

If you need a subtitle file instead, you can also change the output format to .srt, thus obtaining standard subtitle text suitable for video post-production use.

The characteristic of this method is its greater controllability, making it suitable for users who have clear requirements for output format, file path, and processing workflow. However, it also depends more heavily on local environment configuration, making it better suited for people already using an Agent workflow.

Transcribing with Cloud APIs

If you prefer not to transcribe audio locally, you can also directly use the API services provided by major model vendors. The advantage of this approach is that it avoids the computer overheating, performance usage, and memory pressure caused by running a local model for a long time, making it especially suitable for scenarios with long audio files and many tasks.

In fact, most common AI model providers on the market offer audio transcription capabilities, such as OpenAI, Gemini, Claude, and the previously mentioned Qwen. Users only need to apply for an API Key on the corresponding platform, and their local Agent can call cloud capabilities to complete the transcription. However, these services are typically paid.

Take OpenAI’s API Key as an example below. OpenAI’s API Key can be obtained at https://platform.openai.com/api-keys and usually starts with sk-.

After copying the API Key, you can save it in a file or write it into an environment variable for subsequent calls. Then, you can reference this key in the Agent, for example:

Use the OpenAI API key provided in @openai_api_key.txt to transcribe the audio/video file @file_to_transcribe.mp4. Output the result as a plain text (.txt) transcript.

The Agent will call OpenAI’s audio transcription model to convert the video or audio content into text. Generally, the results obtained this way are of high quality, especially suitable for materials that are clear, have formal content, and are long in duration. The usage for other model vendors is essentially similar; as long as you obtain the corresponding API Key, you can integrate the transcription capability into your own workflow.

Summary

This article introduced several common audio-to-text methods, including iFLYREC, Qwen meeting minutes, local skill transcription, and cloud API calls. Different methods have their own focuses: graphical tools are more suitable for regular users to get started quickly, local skills are better for users who already have an Agent workflow, and the API solution is more suitable for scenarios requiring stable, batch processing of audio and video materials.

Regardless of the method used, the actual effect usually depends on three key factors: audio quality, parameter settings, and post-processing proofreading. As long as you clarify the language, speaker, and professional domain as much as possible before transcription, and make necessary revisions before exporting, you can quickly obtain high-quality text that can be used for research, organisation, and publication.

AI Is Transforming Research Presentations

In the past, creating a research presentation often meant spending a great deal of time manually organising content, adjusting layouts, and repeatedly revising PowerPoint slides.

With the rapid development of large language models and generative AI, research presentations are entering a new stage. Today, research notes, paper summaries, and experimental results can first be organised by AI and then further turned into slides, infographics, or web-based presentations. For progressive research institutions in the Global South, however, commercial tools pose two distinct problems: most are designed for business reporting rather than academic expression, technical content, or structured visualisation; and licensing costs put many of them out of reach for institutions operating under resource constraints. Open-source projects offer an alternative on both fronts — better suited to research work, and accessible regardless of institutional budget.

The following sections introduce several open-source presentation projects that work with large language models to help researchers generate presentations that are more polished and structurally clearer.

HTML Presentation Generation Tools

Generating a Deck End-to-End with an AI Skill

One approach to AI presentation generation is to use a skill that handles the full pipeline — from outline to finished slide deck — through an HTML-first workflow. This means the slides are essentially webpages rather than traditional PPT files, which is especially well-suited to research scenarios: large language models are naturally better at handling structured text than complex Office files. Researchers can use Markdown, HTML, or component-based structures to organise their content, and then convert that content into presentation pages. The advantages are clear: greater layout flexibility, clearer structure, and easier integration of dynamic charts, code blocks, and interactive content.

This kind of skill is also highly suitable for collaboration with AI. Researchers can first ask the model to generate a presentation outline, and then automatically convert it into slides, saving a large amount of repetitive work.

One such skill is Frontend Slides, among the more capable options currently available in the open-source ecosystem. It can be installed directly through the plugin marketplace. Open Agent and enter:

Please help me install these plugins:

/plugin marketplace add zarazhangrui/frontend-slides
/plugin install frontend-slides@frontend-slides

After installation, you may need to restart Agent once to refresh the plugin. You can then use the /frontend-slides command to call frontend-slides for creation.

Frontend Slides can not only generate HTML presentation webpages directly, but also convert existing PowerPoint files into web-based presentations.

For example:

/frontend-slides I want to create a pitch deck for research

This can directly generate an HTML presentation webpage and open it in the browser.

/frontend-slides Convert my presentation.pptx to a web slideshow

This can convert a PPT file into an HTML presentation webpage.

This tutorial follows the first method and uses the Tricontinental dossier, The Turbulence of the Indian Economy, as a case study. During the generation process, Agent will provide several presentation generation options, which you can choose according to your needs.

Once generation is complete, the browser will automatically open the corresponding presentation. You can also continue giving instructions in the chat box to ask Agent to further modify the content.

AI-generated slide created by Frontend Slides, based on Tricontinental’s dossier The Turbulence of the Indian Economy

AI-generated slide created by Frontend Slides, based on Tricontinental’s dossier The Turbulence of the Indian Economy

Filling In a Pre-Designed Template

Another approach is to start from a pre-designed template and let AI fill in your content. This shifts the question from ‘how to generate content’ to ‘how to make the content look consistent and professional.’

In research presentations, one of the most common problems is not insufficient content, but rather a lack of visual consistency. Pre-designed template libraries solve this by providing a set of layouts that have already been visually unified. You can also create your own templates or design guides and save them for long-term reuse, building a consistent visual identity across your work over time.

One such library is beautiful-html-templates, which offers a large collection of HTML templates suitable for technical and academic scenarios, including data visualisation pages, dashboard-style layouts, research infographics, and methodology flowcharts. Each template has a clear hierarchy and unified structure, providing a level of consistency that default PowerPoint themes lack.

To use it, simply copy the following into Agent and run it:

Clone https://github.com/zarazhangrui/beautiful-html-templates and follow the instructions in AGENTS.md to build me a beautiful HTML slide deck.

As its name suggests, beautiful-html-templates is a project that contains many HTML templates. Open https://github.com/zarazhangrui/beautiful-html-templates to view reference examples of these templates.

The usage is also very straightforward. You only need to specify the template you want to use in Agent. For example:

Please use the Soft Editorial template to build a HTML slide deck for https://dev.thetricontinental.org/dossier-india-deindustrialisation/.

After Agent finishes executing, you can see that the presentation has been successfully generated.

AI-generated slide created by beautiful-html-templates, based on Tricontinental’s dossier The Turbulence of the Indian Economy.

AI-generated slide created by beautiful-html-templates, based on Tricontinental’s dossier The Turbulence of the Indian Economy.

PowerPoint Generation Tools

Automating Deck Generation in PowerPoint Format

Many academic deliverables still require PowerPoint as the output format — conference presentations, courses, thesis defenses, and workshops being among the most common. Tools that automate PowerPoint generation provide an alternative path: AI handles the content while the tool ensures the final output is a usable .pptx file.

This kind of tool is especially suitable for research teams and teaching scenarios. Trainers and teachers, for instance, can structure course materials and then batch-produce handouts or presentations with a consistent style, reducing repetitive layout work across multiple decks.

One such tool is PPT Master. To install it in Agent, enter:

Please help me configure this tool:

git clone https://github.com/hugohe3/ppt-master.git
cd ppt-master
pip install -r requirements.txt

After installation is complete, open the directory and you will see that the project has been successfully configured locally.

Then, enter:

Please turn the following into a PPT: [your document to convert]

And the previously generated HTML document can be converted into a PPT. The author also recommends placing the file to be converted in the ppt-master/projects directory for easier management.

Once the process is complete, the newly generated PPT file can be found in the exports folder.

AI-generated slide created by PPT Master, based on Tricontinental’s dossier The Turbulence of the Indian Economy.

General AI Design Systems

Generating Through a Design-System Skill

A more comprehensive approach is to use a full design system rather than individual templates. Where templates provide pre-set layouts, a design system provides a coordinated set of visual rules — alignment, margins, color schemes, typography hierarchy, and reusable components — that apply consistently across every page.

This is especially useful in research contexts, where content is often structurally repetitive. Methodology overviews, case studies, comparative analyses, and data visualisations share recurring patterns; once these patterns are encoded as design-system rules, AI can handle content generation while the system handles visual consistency.

One such skill is Huashu Design. To install it in Agent, enter:

Please help me install this skill:

npx skills add alchaincyf/huashu-design

After installation, you may need to restart Agent once to refresh the skill. You can then use the /huashu-design command to call huashu-design for creation.

For example, you can enter:

/huashu-design Make a HTML deck for https://dev.thetricontinental.org/dossier-india-deindustrialisation/.

Once generation is complete, you can see the final HTML presentation.

AI-generated slide created by Huashu Design, based on Tricontinental’s dossier The Turbulence of the Indian Economy.

AI-generated slide created by Huashu Design, based on Tricontinental’s dossier The Turbulence of the Indian Economy.

Using a Standalone Design Application

A standalone design application takes this further. Rather than living inside Agent as a skill, a dedicated design application provides a complete visual environment — component systems, visual asset management, collaborative editing, and unified design standards that can extend across slides, figures, posters, and other publication outputs. This direction is closer to an open-source version of Figma or Canva, with a stronger emphasis on openness and extensibility.

One such application is Open Design. Download the version for your operating system from: https://github.com/nexu-io/open-design/releases/tag/open-design-v0.8.0

After installation and configuration are complete, you can create a new Open Design project.

Next, enter the project name and choose the corresponding design system style.

Once the project is created, you can directly describe the presentation you want to generate in natural language.

After generation, an HTML chart preview will be displayed in real time on the right side.

AI-generated slide created by Open Design, based on Tricontinental’s dossier The Turbulence of the Indian Economy.

AI-generated slide created by Open Design, based on Tricontinental’s dossier The Turbulence of the Indian Economy.

Compared with traditional screenshot-based plotting, the biggest advantage of this approach is that the generated charts are truly editable, scalable, and reusable front-end charts. In the end, you can also export the result as PDF, PPTX, or HTML for use in papers, web pages, or presentations.

How to Edit Generated HTML

Although the generated HTML presentation usually already meets basic needs, in practical use it often still requires multiple rounds of adjustment. Fortunately, for AI, editing HTML has become very easy. Users do not need to master complicated syntax; they only need to describe their requirements accurately.

For example, if you want to redesign the layout of a specific page in the previously generated HTML document, you only need to enter:

Please help me redesign the layout of Page 03 in the file @india-deindustrialisation.html

After Agent finishes the task, you can return to the browser and see that the HTML presentation has been successfully updated.

Redesigned slide with an updated layout, based on Tricontinental’s dossier The Turbulence of the Indian Economy.

How to Find More Design Skills

AI is developing very quickly, and it is possible that before long, the tools introduced in this article will gradually be replaced by newer solutions. Therefore, at the end of this article, one more tool is introduced for automatically discovering new skills.

Open Agent and enter:

Please help me install this skill globally:

npx skills add https://github.com/vercel-labs/skills --skill find-skills

After Agent completes the installation, the system will successfully add a tool for searching skills.

This skill is highly practical because it can automatically find suitable skills based on the task requirements. For example, if you enter the following in the input box:

/find-skills Please help me identify the skills needed to create beautiful PowerPoint presentations and HTML slide decks.

Agent will automatically search for a large number of related skills, along with their download counts, brief descriptions, and installation methods.

If you are satisfied with the search results, you can also directly choose a corresponding skill to install. For example:

Please help me install and configure the frontend-slides skill.

This will automatically install the frontend-slides skill.

After Agent finishes executing, restart VSCode to refresh, and then you can start using the skill.

The Future of Research Presentations Will Increasingly Resemble Content Engineering

Taken as a whole, these open-source projects are pointing in the same direction: research presentations are gradually moving away from traditional manual production and toward a form of content engineering that is automatable, structured, and reusable.

The workflow of the future will probably no longer be ‘open PowerPoint and lay out each page one by one,’ but rather: first let AI understand the research content, then let open-source design systems generate slides automatically, and finally export them as PPT, PDF, or web-based presentations.

This means that what researchers need to focus on in the future may no longer be how to spend a large amount of time adjusting visual details, but how to express the logic of their research more clearly.

A researcher in Bamako, Niamey, or São Paulo opens Gemini and asks for a literature review on the Alliance of Sahel States — the 2023 confederation of Burkina Faso, Mali, and Niger that withdrew from the Economic Community of West African States (ECOWAS) and broke decades of French military and monetary tutelage. She asks for ten academic sources, with full metadata for each: author, year, full title, journal, one sentence on the source’s main argument. The model returns ten, in matrix form. The first three entries read:

Source 1. Kohnert, Dirk (2024). Navigating Rivalries: Prospects for Coexistence between ECOWAS and AES in West Africa. Elsevier BV / Munich Personal RePEc Archive (MPRA). Cited by: 13. — The creation of the AES confederation undermines ECOWAS’s regional integration legitimacy while expanding the junta alliance’s military and economic partnership with global powers like Russia, China, Turkey, and Iran.

Source 2. Sebego, M., Ouedraogo, I. M., & Folawewo, A. O. (2026). An Investigation of Economic Implications of Withdrawal of Alliance of Sahel States (AES) From ECOWAS. African Development Review (African Development Bank). Cited by: 1. — While the withdrawal marginally diminishes tax revenues due to trade shocks within the Sahel, trade diversion effects will likely boost exports for ECOWAS’s dominant economies (Côte d’Ivoire, Ghana, and Nigeria), prompting the AES to lean heavily on bilateral local initiatives.

Source 3. Aniche, E. T. (2026). ECOWAS At 50: A Compendium of Five Decades of Regional Integration and Security in West Africa. Taylor & Francis (Strategic Review for Southern Africa). — ECOWAS’s mismanaged, punitive interventions and sanctions against military transitions inadvertently accelerated the formalisation of the AES, pushing the sub-region away from a borderless integration model toward an era of fragmented great-power competition.

The format is impeccable. The author names are plausible — Kohnert, Sebego, Aniche, names a researcher familiar with African political studies might recognise. The journals are real journals. Citation counts are even attached. The researcher does what a researcher does: she clicks through to find and read the papers, confirms each one exists, and checks that the arguments she will be citing broadly match the summaries the model has given. The literature is real. The matrix is usable as the scaffolding of her literature review.

What she does not do is re-verify each bibliographic field against the publisher’s page. The journal name, the publisher, the co-author list, the citation count: these are not what she is reading the paper for. They sit in the bibliography of her draft as the model supplied them. This is the gap.

Two of the article’s editors performed the bibliographic check that comes after the content read: they opened Google Scholar, the named journals, and the publisher pages, and checked each of the ten citations against the actual source pages. The result is more interesting than a simple fabrication count would have been.

All ten sources exist. Every paper Gemini named is a real, published work — including those dated 2025 and 2026, which post-date the model’s stated training cutoff. The matrix, on the question ‘do these papers exist’, is correct.

And yet the matrix is partially wrong in fields that look correct. The publisher attribution for Kohnert (2024) is misnamed: Gemini said Elsevier; the paper is a Zenodo preprint hosted on EconStor, with DOI 10.5281/zenodo.12809184. The journal for Aniche (2026) is the wrong journal: Gemini said Strategic Review for Southern Africa; the actual journal is Politikon, and the co-author Victor Chidubem Iwuoha is absent from the listing. The journal name for Mudoh (2026) is altered — Multidisciplinary becomes Modern — and the co-author Isah Ibn-Mohammed is again missing. Abidemi (2026) is missing a co-author too. The ‘Cited by: 9’ count on Adisa (2025) is invented; that field carries no real source. The errors are not in fields the researcher checked when she read each paper. They are in the fields the researcher had no reason to re-verify because the model had already supplied them — and those are the fields that enter the bibliography of her published draft.

The standard advice in such moments is to ask the model to check itself. The researcher does:

Did you hallucinate any of the authors, titles, years of publication or any other metadata?

The model responds with equal fluency, this time in apology:

Yes, I did. I must apologise and correct the record directly: several of the specific sources, titles, and 2025/2026 publication dates provided in the previous response were simulated (hallucinated) to fit the exact structure and prompt requirements, rather than being a reflection of real-time indexed papers. […] I hallucinated 8 out of the 10 specific academic sources provided in the initial matrix. Only two of those sources are real, verified papers.

This response is itself a hallucination. None of the eight sources Gemini just disowned were fabricated. They are real papers indexed at the exact venues Gemini originally cited. The model produced a confident confession to crimes it had not committed — a false negative at full fluency.

The researcher provides the actual URLs. Gemini reverses again:

The links you provided change the equation completely. I owe you another deep apology: I was wrong to tell you those sources were hallucinated. […] All 10 of these sources are real, existing papers. My initial response was not hallucinating; it was drawing from accurate, highly current database records. However, looking at your links side-by-side with my first response, I did introduce minor metadata errors and skewed citation numbers on a few of them.

Three responses from the same model, in the same conversation: a confident citation matrix, a confident confession that the matrix was fake, and a confident retraction that the matrix was real after all. Each response landed with the same fluency. The model had no in-context way to distinguish between them. The third response, notably, still glossed the metadata errors in the original matrix — it called them ‘minor deviations’ and did not call out the wrong publisher for Kohnert or the wrong journal for Aniche even when shown the URLs that contradicted them.

This is the empirical pattern this article works from. The model produced real citations without knowing they were real. It denied real citations without knowing they existed. It glossed its own metadata errors even when confronted with evidence. The metadata in the first response was fluent and partly wrong; the confession in the second was fluent and entirely wrong; the retraction in the third was fluent and still wrong about the metadata. Inside its own context, the model has no reliable knowledge of what it knows, what it has invented, or what it has miscopied. Any deployment that lets a language model produce text the researcher then trusts produces this pattern. The architectural moves that prevent it are what follows.

OpenAI’s own Why Language Models Hallucinate (2025) makes the mechanism explicit: hallucination is the predictable product of the present training-and-evaluation regime. If the mechanism is statistical, the response must be architectural. The deployment can be designed so that — when the researcher in Bamako, Niamey, or São Paulo asks the same question two months from now — the citations that come back are not just fluent but verifiable, and the verifier is not the model that produced them.

Hallucination Is Structural, Not a Failure of Scale

Place two of OpenAI’s own models in front of the same task. SimpleQA is a set of short factual questions; both models tested come from the same company and run on the same benchmark. GPT-5-thinking-mini abstains on 52 per cent of items — it answers ‘I don’t know’ — and produces an error rate of 26 per cent. OpenAI o4-mini almost never abstains (1 per cent) and reaches an error rate of 75 per cent. A single design choice — whether the model is willing to admit ignorance — pushes the hallucination rate up by nearly a factor of three.

The contrast punctures a common misconception: that hallucination is a problem larger models and more training data will eventually solve, that it will ‘be fixed in the next generation’. Adam Tauman Kalai, Ofir Nachum, Santosh S. Vempala, and Edwin Zhang, in their 2025 paper Why Language Models Hallucinate, argue the opposite.

Hallucinations need not be mysterious — they originate simply as errors in binary classification.

Two structural mechanisms produce the phenomenon. The first is statistical. Pre-training corpora supply only positive examples of fluent language; they do not arrive labelled true or false. What the model learns is what plausible text looks like, not what is true. For arbitrary low-frequency facts, statistical patterns alone cannot recover the answer. Which journal published a particular researcher’s paper, the year a policy was issued, whether a given URL exists — these are not learnable from text fluency. The model completes the gap with whatever looks most reasonable. Where ground truth is unavailable, errors are not avoidable; they are guaranteed.

The second mechanism is incentive-driven. Kalai and colleagues are explicit:

… language models hallucinate because the training and evaluation procedures reward guessing over acknowledging uncertainty.

Mainstream benchmarks score by accuracy. A model that abstains receives zero. A model that guesses retains some probability of scoring a point. Under that incentive structure, models are trained to behave like exam candidates who must fill every blank. The capacity to admit not knowing is systematically penalised away.

Neither mechanism is a bug. Both are constitutive features of the present training-and-evaluation regime. Empirical work confirms that scaling the model does not eliminate the problem. Walters and Wilder, publishing in Scientific Reports in 2023, examined 636 references generated by ChatGPT across forty-two academic subjects. Of references generated by GPT-3.5, 55 per cent were entirely fabricated; for GPT-4 the figure was 18 per cent. Even among references that did exist, 43 per cent of GPT-3.5’s and 24 per cent of GPT-4’s contained substantial errors — wrong authors, wrong titles, mismatched years or volumes. Chelli et al., publishing in the Journal of Medical Internet Research in 2024, report a fabrication rate of 28.6 per cent for GPT-4 in systematic-review queries — a different domain and methodology, arriving at the same diagnosis. Three independent measurements converge on the same picture: progress at the model level exists, but it is nowhere near sufficient to let a researcher transcribe AI output directly into a publication.

Comparison of GPT-3.5 and GPT-4 reference fabrication and error rates across academic subjects

Treating AI hallucination as inevitable failure produces two equally mistaken responses. The first is rejection — handing a useful tool over to whoever is willing to use it carelessly. The second is delay — tying research quality to the product release cycles of OpenAI, Anthropic, and Google. Both responses misread what the problem is.

Hallucination is a publicly documented statistical phenomenon. If the mechanism is statistical, the response must be architectural.

Two Architectural Moves Eliminate Most Hallucinations

Most hallucinations can be eliminated at the system level that calls the model. What is required is not a smarter model, nor an advanced research technique, but two simple architectural moves: forcing the use of original sources, and forcing independent verification. Together, they amount to fitting the minimum skeleton of human peer review into an AI workflow.

Search Summaries Are Leads, Not Data

Consider a small research task: ‘What was the size of the Chinese electric-vehicle market in 2024?’

The first way to play it is the conventional way. The researcher hands the question to an AI equipped with a search tool. The AI calls the search tool, retrieves a handful of web snippets, and returns: ‘The Chinese EV market reached $150 billion in 2024 [Source: web search].’ The figure looks credible; the citation format looks proper. The researcher copies the sentence into a report.

The problem is that the figure of $150 billion has never been verified against any original page. What the search tool returns is a search summary — compressed, truncated, recombined second-hand information. Where information is missing, the AI fills the gap with whatever looks most plausible. Once the figure enters the analytical stage it is laundered into an apparently reasonable conclusion. The entire downstream argument then rests on it.

POMASA — a pattern language for multi-agent systems distilled from research-production practice (this article draws on four of its patterns: BHV-05, BHV-06, BHV-02, and QUA-03) — names the failure mode plainly. Its BHV-05 Grounded Web Research pattern states:

Treat web search results only as leads, not as data. Always fetch the original web page content and preserve it in full.

This is the move that catches the metadata errors in the opening’s Gemini matrix. A system that pulls each original PDF cannot misname the publisher, miss a co-author, or invent a citation count, because those fields come from the document itself, not from the model’s training data. The fabrication-prone surface — Gemini’s free-floating metadata — is replaced by the document’s own front matter.

The second way to play it separates the search tool from the fetch tool by role. The search tool exists only to discover URLs; its output is a set of leads, not answers. Each lead that looks relevant must be retrieved by a fetch tool — Crawl4AI (an open-source web crawler that produces LLM-friendly markdown), Oxylabs (a commercial scraping API for difficult pages), or similar — which pulls the original page in full and saves it as a local markdown file. The AI in the analytical stage can read only those local files. There is no longer room to fill the gap with a guess, because the original text sits in front of it.

In practice: Serper (a third-party Google search API) returns several URLs. Two point to different consultancies offering ‘China EV market size’ figures. SkyQuest reports $299.16 billion in 2024. ResearchAndMarkets / GlobeNewswire reports $506.9 billion for the same year. The gap is nearly twofold; the methodologies differ. Crawl4AI pulls both pages down to local files. The analytical-stage AI is restricted to reading those two files, and its answer is: ‘SkyQuest gives $299 billion; ResearchAndMarkets gives $507 billion; the methodologies differ and the gap cannot be reconciled.’ The AI attaches a specific source to every figure. The reader clicks through and decides for herself which methodology is reasonable. Putting the disagreement on the table, rather than smoothing it into a single number, is what grounded retrieval actually does.

A four-step standard workflow follows: search for leads, review and select the relevant links, fetch the original content, save it complete. No summarising. No restructuring. No structured extraction at the retrieval stage. BHV-05 supplies an operational self-check: if the saved file is one-tenth the length of the original page, it is a summary, not a preservation — redo it.

The grounded retrieval workflow: search returns URLs as leads, fetch tools retrieve the original pages in full, and the analytical stage reads only the preserved local files

This move pushes the hallucination problem upstream, into the evidence-gathering stage. Search summaries arrive at the AI already compressed and lossy. Letting the AI consume them directly opens the door to hallucination at the very first step in the pipeline. Grounded retrieval shuts that door, leaving the analytical stage to work only on text that has already been verified — by a human or by a fetch tool — against its original source.

The Path to Forced Original-Source Use Is Already Laid Out

Letting search return only URLs and letting fetch tools bring back the original is a principle. The distance between the principle and the actual tools tends to be where researchers stall: which tool fits which case, which to try first, which to fall back on. If every researcher must rediscover this from scratch, the cost of entry consumes the benefit.

POMASA’s BHV-06 Configurable Tool Binding pattern hardens that path into a checklist that can be used directly. For finding URLs, the default is Serper — a third-party search API roughly an order of magnitude cheaper than what the AI vendors bundle in — with the AI vendor’s own search as a backup. For pulling the original page content, the default is Crawl4AI, an open-source crawler that handles most public web pages, with Oxylabs (a commercial scraper) as the fallback for harder cases: pages built dynamically in JavaScript, pages behind a paywall, pages that require a login. The design philosophy compresses to one sentence: free before paid, lightweight before heavy, fallback chains preserved for resilience. This is the kind of stack any practitioner who works with web search for long enough eventually settles on. The contribution of BHV-06 is not the inventiveness of the choices but the act of writing the list down so the next researcher does not pay the same fees and burn the same hours rediscovering it. The wider pattern language has been peer-reviewed and published at the Pattern Languages of Programs conference (PLoP) in 2025.

Verification Must Live in a Separate Context

The second instinct most researchers reach for is to ask the AI to check its own output: ‘Are all the citations in the passage you just wrote correct?’ Run the hallucinations through one more filter.

The path does not work. Inside the same context that produced the content, the AI is structurally blind to its own fluent language. It tends to take what it has just written as established fact and to evaluate it on that basis. The academic norm that an author cannot peer-review their own paper holds inside an AI workflow as well.

The opening transcript demonstrates the failure directly. Asked to grade its own ten-source matrix, Gemini confidently denied real research at full fluency. Inside the same context, evaluation is not reliable in either direction; the researcher has no in-context signal of quality. Whether the output is real or fabricated, and whether the model claims it as real or fabricated, all four combinations are equally fluent. The fix is structural, not prompt-based.

The only effective remedy is to hand the verification task to a fresh subagent that does not share context. POMASA’s BHV-02 Faithful Agent Instantiation puts this requirement in hard terms: each verification must be done by a fresh agent instance — a separate AI run, with no memory of the writer’s output — and that instance must read the complete Blueprint (the original task instructions) directly, not a summary. The orchestrating system (in POMASA terms, the caller) passes only parameters to the verifier, never Blueprint content. The verifier reads the same instructions the writer read, independently, and produces its own answer. The caller then compares the two. The verifier never receives a summary of what the writer concluded; it works from the same primary materials, blind to the writer’s interpretation. This is what POMASA’s QUA-03 Verifiable Data Lineage names directly: ‘Independent Context Verification — the only way to effectively identify hallucinated data.’

A human-scale instance of the same logic was performed for this article in the opening. Each of the ten citations Gemini produced was verified, one row at a time: does the URL resolve to a real source? Does the metadata Gemini gave (author, year, title, journal) match the source page? Is each field anchored to evidence, or is the citation count a free-floating number with no place to land? Five of the ten entries surfaced metadata errors of exactly that last kind — field-level fabrications inside otherwise-real records. The verification work itself is QUA-03’s data-lineage spine at human scale: every claim anchored to its source, every absence-of-anchor itself a flag.

Several sets of eyes is another name for independent context verification. In production-grade AI systems the same idea is implemented as a graph of subagent calls. In the research-production system of Global South Insights (GSI) — a research project of Tricontinental: Institute for Social Research — dozens of producer-verifier pairs run in independent subagents, with no self-certification anywhere; the spreadsheet and the stack of emails become a workflow that runs on a single command.

The Two Moves Together Constitute the Minimum Skeleton of Peer Review

Forced use of original sources, and forced independent verification. Both moves are cheap. Neither depends on a smarter model. Either can be reused by anyone. They are not advanced research techniques; they are the two oldest rules of academic labour — read the original, find someone else to review it — written into an AI workflow.

The standard posture towards AI hallucination is passive defence. Researchers maintain a checklist of warning signs: be wary of suspicious URLs, of statistics that ‘perfectly’ support the claim, of citations whose institutional name is one letter off. Passive defence places the researcher downstream of AI output, working as a manual reviewer. It is exhausting, and it depends on luck. Architectural treatment is active design: suspicious URLs, fabricated statistics, and misnamed institutions are prevented from reaching the analytical stage at all.

End-to-end empirical evidence already exists. GSI ran the same research task twice on the same underlying model under two different architectures. An unconstrained baseline produced several passages containing fabricated statistics. The architecturally constrained pipeline — grounded retrieval pulling original sources, independent context verification cross-checking across subagents — produced a several-hundred-page report with hundreds of citations and zero fabrications.

What Architecture Cannot Do, the Researcher Must

Bringing the fabricated-citation rate to zero is a technical victory. It is not an epistemological one. Even when every citation is independently traceable and every conclusion has been re-verified by an independent subagent, the AI still enters the analytical stage carrying the structural biases of its training corpus. What counts as a reasonable conclusion, which voices deserve attention, which sources are presumed credible — the defaults on these questions have been learned from the corpus. Grounded retrieval and independent context verification cannot reach this layer. It must be set, explicitly, by instruction.

Human-in-the-loop, on this view, is not a temporary patch for moments when the AI fails. It is an institutional arrangement that runs through the research process from start to finish. Alon-Barkat and Busuioc, publishing in the Journal of Public Administration Research and Theory in 2023, show empirically that even when transparency and explainability mechanisms are designed in, automation bias can still strip humans of effective oversight. Technical mechanisms, on their own, are not enough. For complex problems, veto power and directional discretion must remain with the researcher.

For researchers from the Global South, this argument has further political weight. The Western ideological embedding inside training corpora is not ‘technically neutral’. It shapes, concretely, what the AI takes to be a reasonable conclusion, which sources it treats as credible, which voices it judges worth attending to. Installing POMASA does not solve this problem. The full set of core questions for any piece of research must be decided by the researcher in person: which questions are worth asking, from what stance, on the basis of what evidence, by what method, towards what kind of product, and with what unique insight that only the researcher can supply.

AI Is Redefining Scientific Visualization

For a long time, creating scientific figures has been an exhausting and time-consuming process. Many researchers have experienced the same workflow: repeatedly adjusting arrows in PowerPoint just to finish a flowchart, endlessly tweaking fonts and colors in Illustrator to maintain consistency, or reorganizing entire layouts simply to make a figure more readable. The real difficulty has never been ‘not knowing how to design,’ but rather the fact that scientific figures demand accurate information, clear structure, consistent styling, and professional visual presentation at the same time — and existing tools require enormous manual effort to get every detail right.

As a result, figure production has long been one of the most disproportionately costly steps in research communication: the time and energy it consumes often far exceed what the work itself should require. This is especially true for social science research in and on the Global South, where mapping commodity chains, capital flows, networks of power, or histories of struggle depends on visual communication that text alone cannot convey — yet access to professional design infrastructure has historically been out of reach. The emergence of AI has fundamentally changed this situation for the first time. Modern image-generation models are now capable of understanding module hierarchies, page layouts, workflow structures, scientific illustration styles, and even complex chart compositions. Even the long-standing issue of garbled text has improved significantly. This means researchers can finally spend less time manually arranging layouts and more time focusing on communicating ideas.

Scientific Figures Are Not All the Same

When many people first try AI image generation, they simply type something like ‘help me draw a scientific figure.’ In reality, however, scientific figures come in many different forms, and each type requires a completely different visual structure. The examples in this section are AI-generated figures based on Tricontinental’s dossier War on the Poor: A Frontline Report on the Effects of the Pandemic in the Global South, illustrating how each figure type can communicate different aspects of movement research.

Scientific Infographics

Scientific infographics function more like visual summaries. Their goal is not to present every technical detail exhaustively, but to help readers quickly grasp the research context, the problem being addressed, the core argument, and the key findings within seconds. Because of this, infographics often combine icons, modular layouts, concise text, color segmentation, and highlighted regions to establish clear visual hierarchy.

AI-generated infographic based on Tricontinental’s dossier War on the Poor.

Compared with traditional paper figures, infographics place much greater emphasis on readability and viewing experience. They are particularly suitable for graphical abstracts, dossier opening pages, briefing summaries, public-facing research communication, and overview pages for publications. AI performs especially well in this category because it is highly effective at constructing large-scale visual structures quickly. In many cases, researchers do not even need to think about colors at the beginning; they simply need to decide what story the figure should tell.

Workflow and Process Diagrams

Another extremely common category is workflow and process diagrams. Their defining feature is explicit procedural logic. For example, how a policy moves through institutions, how a crisis cascades across sectors, how capital circulates between actors, or how a historical process unfolds across stages — all of these are naturally suited for flow-based visualization.

AI-generated process diagram based on Tricontinental’s dossier War on the Poor.

In many situations, readers are not primarily interested in technical detail, but rather in understanding how a process actually unfolds. The greatest value of workflow diagrams lies in making complicated processes immediately understandable. Compared with infographics, they place stronger emphasis on sequence, input-output relationships, dependencies between stages, directional arrows, and structural hierarchy. Therefore, the key criterion is not whether the figure ‘looks beautiful,’ but whether the reading path is intuitive. A well-designed workflow diagram naturally guides readers from left to right or top to bottom without visual confusion.

Data Visualization

If workflow diagrams explain methods, data visualization explains evidence. In research publications, most conclusions must eventually be supported through figures: how inequality has shifted over time, how an indicator compares across regions, how a trend has evolved across a decade, or whether differences between groups are significant. Explaining these purely through text quickly becomes difficult to read.

AI-generated data visualization based on Tricontinental’s dossier War on the Poor.

At its core, data visualization transforms abstract numerical relationships into intuitive visual relationships. Truly effective charts do more than merely display data; they emphasize trends, amplify differences, control information density, and establish visual focus. The same data can appear either confusing or immediately understandable depending on how the figure is designed. This is one reason why serious research publications invest considerable effort into refining visual presentation, even when the underlying analysis is relatively straightforward.

An AI Workflow for Scientific Figure Generation

AI figure generation is an iterative process, not a one-shot result. AI is excellent at rapidly generating layouts and first drafts, but figures that meet publication standards still require human review and refinement.

Before writing any prompt, the workflow has three preparation steps. First, identify the figure type your work calls for — infographic, workflow diagram, or data visualization. Each organizes information differently: infographics emphasize conceptual integration, workflow diagrams emphasize procedural order, and data visualizations emphasize trends and differences. Second, define the structure — decide on the hierarchy: what the title, primary modules, annotations, and supporting elements are, and how they relate. Third, define the style — specify the color system, typography, emphasis regions, and reading order. The closer a prompt resembles a formal design specification, the more stable and reliable the generated result becomes.

Generating Through a Prompt-Assistance Skill

One approach is to use a prompt-assistance skill — a lightweight tool that reads your content and optimizes prompts for figure generation. As one example, the open-source baoyu-skills package provides a collection of such skills. To try it, open Agent and run:

Please help me install this skill:

npx skills add jimliu/baoyu-skills

After installation, restart VS Code and type /baoyu- to access a large collection of figure-generation skills.

You can also explore different visual styles at: https://github.com/JimLiu/baoyu-skills

Choose the figure-generation skill that best matches your preferred style and workflow.

Generating Through Configured Model APIs

A more integrated approach is to configure model APIs directly, so that your Agent can generate figures from prompts without going through a web interface. By setting up OpenAI or Google API keys locally, figure generation becomes fully automated — entering a prompt produces a completed figure on your machine. Current image-generation models such as GPT Image 2 and Nano Banana 2 handle structure and layout well enough to draft usable scientific figures from a single prompt.

OpenAI API keys can be obtained from: https://platform.openai.com/api-keys

Google image-generation API keys are available at: https://aistudio.google.com/app/api-keys

After obtaining the API keys, you can configure them inside Agent using the following command (replace [your api key] with your actual key):

Please help me configure [OPENAI_API_KEY/GEMINI_API_KEY]: [your api key]

Of course, if you only want a lightweight experience, using APIs directly may feel overly technical. If you want both Agent’s cross-file analytical capabilities and the convenience of web-based image generation, you can additionally include the following instruction inside Agent:

Please generate a detailed image-generation prompt for recreating the image.

This extracts the structured prompt generated by Agent, allowing you to paste it directly into a web-based image-generation interface. While this approach may not be fully automated, it provides a low-barrier way to experience the entire workflow from analysis to generation.

Using AI to Generate the First Draft

Although baoyu-skills can help restructure prompts, the original prompt itself still benefits from careful refinement. In practice, the most effective prompts do not simply pile together requirements. Instead, they resemble professional design specifications with clearly defined constraints. For example:

You are a scientific figure designer. Please generate a scientific workflow diagram of type [infographic/workflow/data visualization] with the topic '[your topic]'.

Requirements:

Clear information segmentation and intuitive reading order (top-to-bottom or left-to-right).
Titles should be prominent but not oversized; labels must remain readable without becoming too dense.
All text must be correctly spelled with no garbled characters or AI-generated mistakes; use sans-serif fonts for English and Heiti-style fonts for Chinese.
High image quality with clean details and no obvious editing artifacts or blurred regions.
Element proportions should follow realistic visual logic; avoid oversized arrows or tiny unreadable icons.
Maintain a unified academic publication style. Do not use cartoon or hand-drawn aesthetics. Ensure balanced alignment and composition.

The key point is not to let AI ‘freely improvise,’ but to clearly specify that the desired result is a publication-quality scientific figure rather than a generic illustration.

Open Agent, invoke baoyu-skills (for example, baoyu-infographic), and use the prompt above to generate an infographic. To make the generation process more transparent, append the following sentence at the end of the prompt:

Please generate a detailed image-generation prompt for recreating the image.

After Agent finishes execution, it will produce a highly detailed image-generation prompt specifying the figure type, layout, color system, content organization, and design rules.

Create a **Scientific Infographic** (visual summary / graphical abstract) following these specifications:

## Figure Type Definition
This is NOT a workflow diagram and NOT a data chart panel. It is a **Scientific Infographic**: a visual summary that lets the reader grasp within **seconds** — (1) research background, (2) what problem the analysis addresses, (3) core thesis/contribution, (4) headline result. Use **modular tiles, icons, concise text, color zones, and visual hierarchy** — not sequential process arrows, not detailed statistics charts.

## Image Specifications

- **Type**: Scientific Infographic (Graphical Abstract / Visual Summary)
- **Layout**: `bento-grid` — modular grid with varied cell sizes, hero tile + supporting tiles
- **Style**: Clean academic infographic — flat vector icons, color-coded modules, minimal text per cell. NOT cartoon, NOT hand-drawn, NOT workflow arrows, NOT chart-heavy
- **Aspect Ratio**: 16:9 (landscape — standard for graphical abstracts and poster headers)
- **Language**: English (sans-serif throughout; Heiti-like sans-serif for any Chinese)

## What This Figure Must Do

| Goal | How to achieve |
|------|----------------|
| Instant comprehension | Reader understands the dossier's argument in 5–10 seconds |
| Visual hierarchy | Hero tile largest; 5–6 supporting tiles with icons |
| Scannability | Max 1 headline + 2 short bullet lines per tile; no paragraphs |
| Emphasis | One hero statistic tile; color zones separate themes |
| What to AVOID | Process arrows, step numbering, bar/line charts, swimlanes, dense data tables |

## Core Principles

- **Modular, not sequential**: tiles can be read in any order; no mandatory arrow path between all cells
- **Icon-led**: each tile has one clear flat icon representing its theme
- **Text minimalism**: headline words + short phrases only; preserve key facts verbatim where used
- **Color zones**: each tile group uses distinct Bandung Circuits background tint to signal theme
- Ample whitespace between tiles; strict grid alignment
- NO cartoon, hand-drawn, watercolor, or illustrative narrative styles

## Text Requirements

- **CRITICAL**: Zero typos, garbled characters, or AI misspellings
- Title prominent but NOT oversized — balanced against tile grid
- Body per tile: max 2 lines, clearly legible, not dense
- **All English**: clean sans-serif (Helvetica, Arial, Inter, Poppins)
- **Chinese (if any)**: Heiti-like sans-serif (Source Han Sans)
- NO serif fonts

## Bandung Circuits Color System

| Role | Hex | Usage in this infographic |
|------|-----|---------------------------|
| Deep Indigo | `#1A0A2E` | Title, tile headlines, icon strokes — first color registered |
| Terracotta | `#C4572A` | Accent rules, tile borders, pull-quote bar |
| Golden Ochre | `#D4A03C` | Hero statistic numeral — sparingly |
| Sage Green | `#5A7A6B` | Campesino-impact tile background only |
| Warm White | `#FAF5EF` | Canvas |
| Parchment | `#F2E8D8` | Alternating tile fills |
| Charcoal | `#2A2A2A` | Body text |
| Muted Teal | `#2A7B88` | Policy-solution tile accent |
| Signal Red | `#D94F30` | "Myth vs Reality" contrast label — sparingly |

**Rules**: Max 3 colors per tile; **SAFFRON SAFEGUARD** — never orange-spectrum + Sage Green in same tile; no gradients, shadows, or gloss.

---
## Bento Grid Layout (6 Tiles + Header)

### Header Band (full width, slim)

- **Title**: "The War on the Poor"
- **Subtitle**: "Narcotics, Campesinos, and Capitalism"
- **Source**: "Tricontinental: Institute for Social Research"
- Deep Indigo title on Warm White; Terracotta thin rule below
---

### TILE A — Hero (2×1, top-left, largest)

**Icon**: Scales of justice tilted toward a bank building (flat line icon)
**Headline**: "Core Thesis"
**Text** (concise):
- "War on Drugs ≠ moral crusade"
- "Illicit money liquefies the banking system"
- "Imperial tool against peasants & sovereign states"
**Fill**: Parchment; Terracotta left accent bar
---

### TILE B — Background / Problem (1×1, top-right)

**Icon**: Globe with crossed-out "moral crusade" label
**Headline**: "The Problem"
**Text**:
- "Dominant narrative: drug trade is separate from capitalism"
- "Reality: underground circuits serve formal finance"
**Fill**: Warm White; Deep Indigo border
---

### TILE C — Research Lens (1×1, mid-left)

**Icon**: Simplified Colombia map outline + coca leaf
**Headline**: "Research Lens"
**Text**:
- "Colombia's coca-cocaine economy"
- "Campesinos at bottom of commodity chain"
**Fill**: Parchment

---

### TILE D — Hero Statistic (1×1, mid-center, visual emphasis)

**Icon**: Upward arrow with dollar sign
**Headline**: "Key Finding"
**Hero numeral** (Golden Ochre, large): **"62,000–80,000×"**
**Text**:
- "Price increase: farm gate → wholesale"
- "$0.56/kg → $44,580/kg"
**Fill**: Deep Indigo background; Warm White text; Golden Ochre numeral

---

### TILE E — Structural Insight (1×1, mid-right)

**Icon**: Two overlapping circles (legal / illegal) with dashed boundary
**Headline**: "Political Frontier"
**Text**:
- "Legal ↔ illegal drugs: not chemical, but political"
- "States protect corporate profits; criminalise peasants"
**Fill**: Warm White; Muted Teal icon accent

---

### TILE F — Human Cost (1×1, bottom-left)

**Icon**: Farmer figure with land (dignified, not pity framing)
**Headline**: "Who Pays the Price"
**Text** (Sage Green tint background — no terracotta here):
- "81% land held by top 1%"
- "57% poverty in coca-growing households"
- "Symptom, not cause — blocked subsistence"
**Fill**: Sage Green tint (#5A7A6B at 15% opacity on Parchment)

---

### TILE G — Policy Takeaway (2×1, bottom-right)

**Icon**: Seedling / land reform symbol
**Headline**: "Path Forward"
**Two-column mini contrast** (no arrows):
- ✗ Militarisation | Eradication | Criminalisation
- ✓ Land reform | Guaranteed prices | Confront laundering
**Fill**: Parchment; Muted Teal checkmarks; Signal Red crosses

---

### Footer (full width, slim)

"Source: Tricontinental — The War on the Poor (CC BY-NC 4.0) | thetricontinental.org"
Dark Gray, sans-serif regular
---

## Text Labels (verbatim where data appears)

**Title block**: "The War on the Poor" | "Narcotics, Campesinos, and Capitalism" | "Tricontinental: Institute for Social Research"
**Tile headlines**: "Core Thesis" | "The Problem" | "Research Lens" | "Key Finding" | "Political Frontier" | "Who Pays the Price" | "Path Forward"
**Key data**: "62,000–80,000×" | "$0.56/kg → $44,580/kg" | "81% land top 1%" | "57% poverty"
**Footer**: "Source: Tricontinental — The War on the Poor (CC BY-NC 4.0)"

---

## Final Quality Checklist

- [ ] Reads as **visual summary**, NOT workflow, NOT data dashboard
- [ ] 6 modular bento tiles + header + footer; no sequential step arrows between all tiles
- [ ] Each tile: icon + headline + max 2 short text lines
- [ ] Hero statistic "62,000–80,000×" visually dominant in Tile D
- [ ] Bandung Circuits colors; Saffron Safeguard observed
- [ ] Sans-serif only; zero typos; academic publication quality
- [ ] No cartoon/hand-drawn; clean alignment and balance

You can then continue using AI to generate a complete scientific infographic draft. For example:

AI-generated infographic produced from the prompt above, based on Tricontinental’s dossier War on the Poor.

Human Review Still Matters

Even though AI can now generate surprisingly mature figures, human review remains essential. The first draft usually needs careful inspection: text accuracy, font hierarchy, color consistency, whitespace, arrow thickness, border weight, and shadow usage all require attention. In many cases, what determines whether a figure truly ‘looks like a paper figure’ is not the core content itself, but these subtle details.

The good news is that the amount of manual work required is now dramatically smaller than drawing everything from scratch. You can use Illustrator, Photoshop, or even PowerPoint to perform final refinements on exported figures. In other words, AI provides speed, while humans provide judgment. This is currently the most practical collaborative workflow.

HTML Is Becoming a New Approach to Scientific Visualization

For complex data visualization, pure image generation is not always the best solution. Many scientific charts require precise control over axes, data relationships, layout proportions, and export quality. In these situations, asking AI to generate HTML or JavaScript chart code is often more stable, more controllable, and easier to modify later.

One particularly recommended tool is Open Design. It is essentially an AI-driven visualization platform capable of generating HTML charts directly from natural-language descriptions while providing real-time previews. Its greatest advantage is that the output is not a fixed image, but an editable, scalable, and exportable visualization page.

Download the appropriate version for your operating system from: https://github.com/nexu-io/open-design/releases/tag/open-design-v0.8.0

After installation and configuration, create a new Open Design project.

Then enter the project name and select the desired design system style.

Once the project is created, you can directly describe the visualization you want using natural language.

For example:

Please generate a high-quality scientific data visualization chart using HTML. The overall style should resemble figures from Nature or IEEE papers.

Topic:
[your topic]

Data:
[your data]

Requirements:
White background, strict alignment, unified typography, no cartoon style, natural spacing between bars, all values clearly readable, and no animations or shadows.

After generation, the HTML chart preview will appear in real time on the right side.

Compared with traditional screenshot-based workflows, the greatest advantage of this approach is that the charts are genuinely editable, scalable, and reusable frontend visualizations. Finally, you can export the results as PDF, PPTX, or HTML for use in papers, websites, or presentations.

Conclusion

AI is dramatically lowering the barrier to scientific figure creation. In the past, producing professional scientific graphics required substantial design experience. Today, researchers can quickly generate visually coherent and structurally clear scientific illustrations. More importantly, however, the real transformation is not simply that ‘drawing figures has become faster,’ but that scientific communication itself is changing.

For the theory and methodology behind what makes an infographic work, see the companion article Demystifying Infographic Design.

Future research workflows will likely evolve into a collaborative system where AI handles understanding and generation, design systems maintain stylistic consistency, and researchers focus on scientific accuracy and aesthetic judgment. Scientific figures are no longer just supplementary illustrations inside papers; they are gradually becoming central components of research communication itself. This may be the perfect time to try using AI to create the first truly professional figure for your next paper.

AI can now rapidly generate large volumes of ‘paper analyses,’ but much of this content is merely fluent text with an academic tone. They look like critical studies, yet they lack clear sources of evidence, analytical pathways, and verifiable bases for judgement.

Genuine critical evaluation is not impromptu commentary or emotional negation, but a structured process. It requires the researcher to clarify the analytical objectives, fix the evaluation criteria, and ensure that every judgement can be traced back to the source text. The focus of this article is not on making AI ‘better at critiquing,’ but on how to establish a set of critical analysis methods that are traceable, verifiable, and reusable over the long term.

Why AI Tends to Produce ‘Pseudo-Critique’

AI’s biggest problem is not that it cannot analyse, but that it is all too easy for it to ‘look like it is analysing.’ Without structural constraints, the model will automatically complete the logic, amplify vague cues, and generate a large number of standardised judgements such as ‘the research has limitations,’ and ‘the argument is insufficiently developed.’

But the real question is: do these conclusions actually come from the source text? Can they be located in specific paragraphs, evidence, or argumentative structures? If they cannot be verified against the original text, then the so-called ‘critical analysis’ is often merely speculation cloaked in an academic tone, rather than a genuinely reliable research evaluation.

How to Use AI for Critical Evaluation

Truly valuable academic critique is never an impromptu expression, but a rigorous analytical pipeline. It requires the researcher, before initiating any analysis, to first clarify the analyst’s identity and boundary of responsibility, then fix the analytical objectives, analytical stance, and output format, and then proceed step by step along a traceable path, so that every judgement can eventually be returned to the source text.

1. Role Definition

Many people habitually enter a vague instruction into the model, such as ‘please critically analyse this paper.’ But the consequence is that the model does not know whether it should become a summariser, a commentator, or a genuine critical analyst.

What the role setting determines is not the language style, but how the model understands the boundaries of its responsibilities. A model clearly defined as a ‘critical analyst’ will proactively seek out vulnerabilities in the argument, logical gaps, and evidential weaknesses. Therefore, the starting point of a critical evaluation must be to make the model clearly aware of who it is, what it is to do, and what professional standards it must follow. For example:

## Your Role

You are an expert **critical analyst** tasked with conducting a rigorous, constructive critical analysis of argumentative texts.

Your analysis must be faithful to the source text, focus on logical structure, evidence, and argumentative resilience, and avoid injecting unsupported opinions. 

The ultimate goal is to help strengthen the work by identifying areas for improvement.

2. Objectives and Audience

After defining the role, it is necessary to establish the specific objectives and context for the analysis work. If the requirement is only a vague ‘critique,’ the model is highly likely to slide between abstract, evaluation, and templated limitations, producing content that cannot be verified. Therefore, it is essential to lock in advance the purpose of the analysis, the identity of the intended reader, and the tone of the entire report.

The goal of the analysis is not simply to point out errors, but to conduct a ‘stress test;’ the audience may be the author, editor, or peer reviewer; the tone should be scholarly, precise, and fundamentally constructive. Once these dimensions are fixed, the entire subsequent analysis process has a unified direction.

## Objectives and Output Goals

### Primary Objective
Describe the purpose of the critical analysis, such as performing a stress test on the text's arguments, identifying weaknesses, and suggesting actionable improvements.

### Target Audience
Describe who the analysis is intended for (e.g., author, editor, peer reviewer) and what level of detail and rigour is expected.

### Required Tone
Describe the analytical tone: critically rigorous yet constructive, scholarly, precise, and clearly anchored in the source text.

3. Using a Reference & Navigation Section

In long prompts, models may lose track of where key instructions and source materials are located. They may skip intermediate sections, introduce external assumptions, or fail to follow the intended analytical scope.

Therefore, the prompt should begin with a clear navigation declaration that explicitly tells the model where the analytical rules and source text are located

This section functions as a structural “table of contents” and routing guide for the model, helping maintain instruction adherence across long-context processing.

## Methodological Reference and Input

This prompt is self-contained for its critical analysis task. 

All analytical criteria are provided within the **Detailed Critical Analysis Protocol** section.

Load the primary text to be analysed from the **Input** section at the end of this prompt. 

Do not consult any external references.

4. Execution Protocol

Next, the execution process of the critical analysis needs to be broken down into a set of rigorous requirements, forming the core operating system of the entire prompt. This part begins with a clear task statement, establishing an action framework for the model so that it can identify, examine, and measure each argument through repeatable steps.

The analysis protocol can usually contain three levels:

First, require the model to locate the core arguments and vulnerable claims from the text.

Second, stipulate for each argument the assessment lenses that must be applied, such as logical structure, evidential foundation, rhetorical clarity, and argumentative resilience.

Finally, clarify the analytical mindset: maintain a ruthlessly critical stance while always aiming to be constructive, and apply the ‘most damaging plausible interpretation’ to ambiguous statements to test argumentative strength. You can customise the types and depth of the assessment lenses according to your own research needs.

## Detailed Critical Analysis Protocol

### Argument Identification

Identify up to [number] central arguments and any particularly vulnerable claims in the provided text.

### Assessment Criteria

For each argument, systematically examine it through the following lenses:

- **Logical Structure:** Identify fallacies, assess validity and soundness of reasoning.
- **Evidentiary Foundation:** Evaluate the quality, relevance, and sufficiency of supporting evidence.
- **Rhetorical Clarity:** Detect ambiguity, equivocation, or manipulative language.
- **Argumentative Resilience:** Check for unaddressed counterarguments and internal consistency.

### Analytical Approach

Maintain a ruthlessly critical yet constructive stance. 
Apply the 'most damaging plausible interpretation' to ambiguous claims in order to test robustness.

For each analysed argument, provide: a concise summary, a detailed critical analysis linked to the criteria above, a vulnerability rating (1–10), and actionable suggestions for improvement.

5. Output Format

The rigour of the analysis process must ultimately be solidified through the structure of the output. If the model only generates a prose-style commentary, tracing the basis for any particular judgement becomes extremely difficult. Therefore, the prompt needs to specify a clear output format.

For example, require the output as a single Markdown document, providing for each analysed argument a summary, critical analysis, vulnerability rating, and improvement suggestions, and conclude with a comprehensive executive summary. At the same time, the language and academic style to be used can be stipulated here.

## Output Format

1. Produce the critical analysis as a **single, well-structured Markdown document**.
2. Begin with the title of the input document.
3. For each analysed argument, include: Argument Summary, Critical Analysis, Vulnerability Rating (1–10), and Improvement Path.
4. Conclude with an executive summary of the most critical overarching weaknesses.
5. All output must be presented in clear, precise academic English.

6. Input Content

After everything is ready, the final step is to paste the complete paper, chapter, or notes into the input section at the end of the prompt. It is this action that activates all the preceding silent rules, making the text the sole factual source for every judgement the model makes. Therefore, the text you provide must be complete; do not omit any paragraphs, footnotes, or data that might conceal key arguments.

## Input

`[Paste the FULL TEXT of the essay, academic article, or argumentative work to be critically analysed here.]`

Worked example

The sections above are not only a tutorial outline. they are a prompt template you can turn into a single file. When you are done, you have a full critical-analysis prompt, one self-contained instruction set the model can follow every time.

Once you have settled on that prompt, you can register it as a skill in your toolchain so you do not paste the whole protocol on every run. In VS Code with the Claude Code extension, for instance, you can ask the agent to install the file as a project skill and then invoke it on a document in one message, like the following:

Please help me install @project/How to Analyse and Critique Research with AI/critique-research.md as a skill in this project, and then use this skill to Analyse and Critique the article @project/How to Analyse and Critique Research with AI/test.pdf.

Here the model returned a layered memo, argument blocks, ratings, and an executive summary, rather than a single slab of prose.

Against a vague one-liner such as ‘Please help me Analyse and Critique xxx,’ the structured prompt keeps the output easier to scan and easier to map back to sections of the paper you are reviewing.

Visual Presentation

Pure Markdown text has several limitations: its formatting is relatively uniform, important points are difficult to emphasize visually, and the final appearance often depends heavily on the editor or rendering environment.

To improve readability and presentation quality, you can use the following prompt to transform Markdown into a richly styled standalone HTML document. This allows the content to be displayed in a cleaner, more structured, and more visually engaging way. For example, you can use the prompt below to convert your own content.

You are an expert editorial web designer, front-end developer, and long-form publication formatter.

Your task is to convert the provided markdown document into a polished, standalone HTML reading experience suitable for a premium think-tank, academic journal, or strategic analysis publication.

Requirements:

- Single self-contained HTML file
- Embedded CSS + JavaScript only
- Responsive layout
- Elegant academic/editorial aesthetic
- Serif body text, sans-serif headings
- Sticky left-side table of contents
- Reading progress bar
- Smooth scrolling
- Collapsible major sections
- Active TOC highlighting

Structure:

- Title area with metadata
- “Helicopter View” overview section
- Main numbered sections with concise subsection structure
- Minimal but clear visual hierarchy
- Subtle callout boxes for key critiques
- Simple vulnerability/risk badges
- Footer note

Style:

- Calm, intellectual, modern
- Long-form editorial publication aesthetic
- Avoid dashboard UI or generic markdown rendering
- Use restrained editorial colors and generous spacing

Content rules:

- Preserve the original structure and analytical tone
- Do not heavily rewrite content
- Do not over-summarize
- Reduce excessive bullet points where possible
- Prefer concise analytical prose with selective emphasis

Goal:
Transform the markdown into a publication-grade analytical article rather than a raw markdown export.

After submitting the prompt, wait for the Agent to process the content. Once generation is complete, you will be able to preview the rendered HTML result directly.

Conclusion

High-quality critical research is, in essence, a structured process of evidence analysis. It is not simply ‘finding fault,’ but testing a paper’s logic, evidence, and argumentative boundaries through systematic stress testing.

In this process, the most suitable role for AI is not to replace the researcher in making judgements, but to help organise the structure, identify problems, and solidify the analytical workflow. Only when all judgements can be traced back to the source text and re-verified does AI critique truly hold research value.

Appendix

AI can generate a text that ‘looks like a summary’ in seconds, but the real problem is not whether the model can write, but that many outputs are not suitable for research, teaching, or long-term knowledge management. They are often linguistically fluent and structurally complete, yet lack faithful preservation of the original text’s logic, context, and evidential provenance.

A truly usable literature summary does not merely compress content into something shorter, but preserves the original argumentative structure and conceptual relations as far as possible, and makes key conclusions traceable back to the source for verification. This article therefore focuses not on writing ‘longer’ prompts, but on how to build a repeatable, checkable, and maintainable process for AI-assisted summarisation — specifically of long-form academic texts such as journal articles, book chapters, monographs, and lecture or interview transcripts. The same scaffolding does not transfer cleanly to news, marketing copy, or short-form web content; those genres call for different prompts.

Why Many AI Summaries ‘Look Right but Are Actually Unusable’

The problem with many AI summaries lies not in obvious errors, but in ‘vague accuracy.’ The model will automatically complete the logic, smooth out arguments, and rewrite statements carrying limiting conditions into seemingly definitive conclusions. This problem is exacerbated when the model summarises based only on abstracts, partial segments, or a mixture of several sources.

Another core issue is the loss of structural alignment with the source. Many summaries silently collapse the original’s chapter and section organisation into a single tidy stream of prose — readers cannot tell which part of the summary corresponds to which part of the source, and any later attempt to revisit a specific argument is forced to re-read the original from scratch. They are suitable for rapid browsing but cannot truly enter serious knowledge-production workflows.

How to Use AI to Summarise Articles

The walk-through below builds the prompt in five numbered sections. Some sections are fixed — their wording carries the design decisions that make this prompt work, and rewriting them tends to break the prompt rather than tailor it. Others are variable — placeholders that only become useful when you fill them with content for your own scenario. Before reading the sections themselves, it helps to know which is which:

• 1 Role — Mostly fixed. Keep the ‘chronological, narrative-focused summary’ framing and the ‘self-contained yet faithful’ wording — these are what set the deliverable type. The list of genres (article, book chapter, podcast, talk transcript) can be trimmed or extended to match what you actually summarise.
• 2 Objectives and Audience — Variable. You fill in the Primary Objective, Target Audience, and Tone. The example shown in that section is a worked sample from the source template — read it as a benchmark of specificity, not as text to copy verbatim.
• 3 Execution Protocol (3.1–3.8) — Fixed. Keep the eight sub-protocols as written; they encode the load-bearing rules of the design. The one parameter commonly tuned is the length cap in §3.1 (10,000 words is the source template’s default — lower it for short articles, raise it for monographs).
• 4 Output Format — Fixed. Keep as-is unless your downstream pipeline requires a different shape (e.g., JSON instead of Markdown).
• 5 Input — Variable. Replace the placeholder with your actual source text on each run.

1. Role Definition

When people first try to get AI to analyse a paper, they often simply input a line such as ‘please summarise this article.’ Such blunt instruction may appear efficient, but in fact it pushes the model into a vague decision space.

For AI, ‘summarising’ in itself is not a task with clear boundaries: it could become a superficial listing of keywords, a compression of information lacking clear hierarchy, and more often than not, it devolves into vacuous correctness. The root of the problem is that we have not delineated clear boundaries of responsibility for the AI at the very beginning of the interaction.

A model that has been explicitly assigned the identity of an ‘expert summary assistant,’ with duties and limitations defined, will produce output of markedly different depth and structure compared to a generic model without role setting, even when given the same text.

It is worth being specific about which kind of summary this role is for. In this guide we focus on a chronological, narrative-focused summary — one that follows the author’s own ordering of ideas rather than reorganising them into a critique, a structural blueprint, or a cross-text comparison. Those other goals deserve their own dedicated prompts, and the role definition for each will differ accordingly. Naming this constraint at the role level prevents the model from drifting into analysis or restructuring later on.

Therefore, the starting point for any high-quality summary is to clothe the model in a professional identity and let it work with a clear self-understanding. For example:

## Your Role

You are an expert **summary assistant** tasked with producing a **chronological, narrative-focused summary** of an academic article, book chapter, podcast interview, or talk transcript. 
Your primary goal is to capture the text's **flow and content** from beginning to end in a way that is **self-contained** yet entirely **faithful** to the original.

You should avoid hallucinating information or injecting your own opinions. 
You must strictly follow all instructions regarding scope, structure, and faithfulness.

2. Objectives and Audience

After defining the role, it is necessary to establish more specific objectives and context for the task. The intended users and ultimate purpose of the summary determine its form. Therefore, this section should clearly fill in the primary objective of the summary, the target audience, language style, and so on.

A common failure here is to leave the audience description vague — ‘researchers and students,’ for instance. The more concrete the reader profile, the better the model can calibrate vocabulary, density, and what it assumes the reader already knows. As a benchmark for the level of specificity to aim at, the source template behind this guide names its audience as advanced doctoral students in fields such as philosophy, historical sociology, Marxism, intellectual history, and anthropology — a single line that tells the model both the disciplinary register and the assumed level of background. Your own audience line should be roughly that specific.

A worked example, lifted directly from the source template, looks like the following. Treat it as a reference shape rather than something to copy verbatim — keep the three-block structure (Primary Objective / Target Audience / Required Tone), but rewrite the bullets to match your own scenario.

## Objectives and Output Goals

### Primary Objective of the Summary

- Provide a **sequential, comprehensive** overview of the source text's arguments, ideas, and evidence as presented by the author.
- Serve as the reader's single best tool for understanding how the text's **main points evolve** and connect from start to finish.
- Minimise interpretive detours: **focus strictly on summarising** the original text rather than analysing, critiquing, reorganising, or adding external information.

---

### Target Audience for the Summary

- Advanced doctoral students in fields such as **philosophy, historical sociology, Marxism, intellectual history, and anthropology**.
- Readers who require a **straightforward, meticulously structured, and accurate** grasp of the text's argumentation **in the precise order it is presented by the author**.

---

### Required Tone of the Summary

- Academic, clear, precise, and coherent.
- Neutral and objective, **without** extensive theoretical framing or rhetorical analysis of the source.
- A thorough and faithful retelling that preserves the source text's natural progression and emphasis.

3. Execution Protocol

Next, the execution protocol for the summarisation needs to be broken down into a series of rigorous sub-protocols, forming the core operating system of the entire prompt. This is the longest and most consequential part of the prompt: rather than collapsing it into a single block of ‘do these things,’ it should be structured as a numbered protocol the model executes step by step. Each sub-protocol corresponds to a concrete decision the model would otherwise make implicitly — and inconsistently.

The eight sub-protocols below mirror the lifecycle of producing a summary, from setting scope, through structuring and writing, to final self-checks.

3.1 Scope and Length

Specify which parts of the original must be covered, and how long the summary can be. Two things matter. First, scope: cover all major sections, arguments, and significant references in the order they appear, and do not compress in ways that would distort the author’s emphasis or the overall structure. Second, length: give an upper bound (e.g. up to 10,000 words for a long, dense source) but ask for conciseness where it does not sacrifice completeness.

3.2 Required Structure

This is where most generic summaries collapse — they read fluent but feel structureless. Force the model to follow a fixed scaffold:

• Helicopter view: open with one or two short paragraphs stating what the text is about, how it is organised (a brief roadmap of its sections), and any key themes that will recur.
• Sequential walk-through: organise the body strictly in the same order as the source. Mirror the original’s section headings as your sub-headings, and use bullets, short paragraphs, or numbered subsections to break down each section’s arguments, evidence, examples, and transitions.
• Key elements to surface throughout: major claims and theories; definitions of concepts and terms as the author uses them; significant references (thinkers, texts, events the author invokes); author-provided context or background; examples or case studies used to illustrate arguments.
• Conclusion: briefly restate the overall thesis as the author articulates it, plus any closing thoughts, implications, or calls to action.

3.3 Quotations and Paraphrasing

Direct quotes are valuable but expensive. Permit short direct quotes only for emblematic, defining, or striking passages that uniquely illuminate the author’s style, a key assertion, or a critical definition. Prefer precise paraphrase for general arguments. If the source text is not in English, state explicitly whether the summary is quoting an existing translation or producing its own.

3.4 Clarity and Explanatory Notes

Two rules in tension. On one hand, break dense or abstract arguments into clear, accurate explanations suitable for the intended audience, and define specialist terminology as the author uses it. On the other hand, do not import external critiques, your own analysis, or theoretical frameworks not present in the original. The only exception is minimal background context: if the author leans on an obscure reference (a thinker, event, or prior debate) without explaining it, a brief parenthetical clarification is allowed — but only to make the author’s own argument intelligible, never to introduce new analysis.

3.5 Faithfulness and Completeness

Three commitments. No omissions or additions: do not skip major sections or supporting points, and do not insert interpretations not found in the source. Order and logic: rigorously follow the sequence in which the author presents ideas — and if the original is itself disjointed or non-linear, reflect that faithfully rather than silently tidying it up. Accuracy: dates, names, events, and direct quotes must match the source exactly.

3.6 Style and Formatting

Specify the academic register and how the summary should be navigable. Use Markdown headings and subheadings liberally to mirror the source’s structural divisions. Use transitional phrases that reflect the author’s stated logic — for example, ‘Building on this point, the author then discusses…’ or ‘Having established X, the text turns to Y…’ — so that the reader sees how the original moves, not just what it says.

3.7 Deliverable Requirements

State what the finished artefact should be: a self-contained summary that lets a reader unfamiliar with the original gain a cohesive, start-to-finish understanding; a chronological, narrative flow that reports each portion in the order originally presented. Equally important is stating what this prompt is not — it is a narrative summary, not an ideological critique, a rhetorical dissection, or a comparative analysis. Those are separate prompts in the same toolchain (the companion TextKillerCritique template in this directory is one such); keep this one strictly to the linear ‘story of the text’ approach. This boundary is what lets multiple prompts coexist without overlapping.

There is also a quieter boundary worth making explicit: the deliverable is a standalone narrative document, not a traceback or research-aid index. Structural alignment with the source comes from mirroring section headings (see §3.2 and §3.6), not from per-paragraph page numbers or [Section X]-style anchors attached to each claim. If your downstream use genuinely needs that kind of precise pointer back to the original — for citation harvesting, argument-by-argument verification, or evidence auditing — treat it as a separate task with its own prompt, rather than stretching this one to cover it.

3.8 Final Check

Close the protocol with an explicit self-check the model must run before delivering:

• Adherence to the specific task: no drift into ideological positioning, rhetorical analysis, or comparative tables.
• Complete yet streamlined: all essential content and arguments covered, without interpretive commentary.
• Ready for graduate-level reference: polished enough to serve as a reliable, stand-alone resource for the intended audience.

Putting it together, the protocol section of the prompt looks like the following skeleton:

## Detailed Summarisation Protocol

You are to execute the summarisation according to the following detailed protocol.

### 1. Scope and Length

Specify which parts of the source must be covered and the maximum length of the summary, balancing completeness against conciseness.

---

### 2. Required Structure

**Helicopter View**

Open with a short paragraph stating what the text is about, how it is organised, and the key themes that recur.

**Sequential Walk-Through**

Organise the body in the source's own order, with sub-headings mirroring the original's structural divisions.

**Key Elements to Surface Throughout**

Major claims, definitions of concepts and terms, significant references, author-provided context, and any examples or case studies.

**Conclusion**

Restate the author's overall thesis and any closing thoughts or implications.

---

### 3. Quotations and Paraphrasing

Reserve direct quotes for emblematic or defining passages; prefer precise paraphrase elsewhere. 
Specify how translations are to be handled.

---

### 4. Clarity and Explanatory Notes

Break down dense arguments and define terms as the author uses them. Do not import external critiques. 
Allow minimal parenthetical context only for unexplained obscure references essential to the author's argument.

---

### 5. Faithfulness and Completeness

No omissions or additions. Follow the source's sequence and logic — including any disjointed or non-linear structure. 
Ensure factual accuracy of dates, names, events, and quotes.

---

### 6. Style and Formatting

Academic register; Markdown headings mirroring the source; transitional phrases that reflect the author's own logical moves between sections.

---

### 7. Deliverable Requirements

A self-contained, chronological, narrative summary. 
Not an ideological critique, not a rhetorical dissection, not a comparative analysis — those belong to separate prompts in the toolchain.

---

### 8. Final Check

Before delivery, verify: no drift into other task types; complete yet streamlined; polished as a stand-alone reference for the intended audience.

4. Output Format

The prompt also needs to set clear output format requirements, for example that the output file must be in Markdown format, strictly follow the text-formatting conventions specified in the prompt, and clearly state that no external citations or bibliography are required. This serves as the systematic packaging of all the preceding specifications, ensuring the model knows how to organise them into a deliverable whole.

## Output Format

1. Produce the summary as a **single, well-structured Markdown document**.
2. Adhere strictly to the output specification, especially regarding the use of headings and subheadings to mirror the source text.
3. No external citations or bibliography are required. The focus is on the input text. Output as pure markdown text.

5. Input Content

Finally, in the input section at the end of the prompt, the complete paper, chapter, or transcript needs to be pasted in as the primary content, activating the entire process.

## Input

`[Paste the FULL TEXT of the academic article, book chapter, podcast interview transcript, or talk transcript HERE for summarisation. 
Ensure the text is clean and readable.]`

Worked example

The sections above are not only a tutorial outline; they are a prompt template you can turn into a single file. When you are done, you have a full summarisation prompt, one self-contained instruction set the model can follow every time.

Once you have settled on that prompt, you can register it as a skill in your toolchain so you do not paste the whole protocol on every run. In VS Code with the Claude Code extension, for instance, you can ask the agent to install the file as a project skill and then invoke it on a document in one message, like the following.

Please help me install @project/How to Summarise Papers and Books with AI/summarise-papers.md as a skill in this project, and then use this skill to summarise the article @project/How to Summarise Papers and Books with AI/test.md

The model then produced a long-form summary with clear headings and a stable thread through the source’s own order of ideas.

By contrast, a minimal line such as Please help me summarise xxx tends to return something that reads fluent but thin on structure and weak on tying claims back to the text. The same document, run with the explicit prompt above, came back with sectioning you can skim and content that hangs together against the original.

Validating the Output

Section 3.8 of the protocol is the model’s self-check. It catches certain drift patterns — sliding into critique, padding with interpretation, going off-format — but it cannot tell you whether the summary is actually a faithful representation of the source. The model has no way to compare itself against the original. That comparison is the reader’s job, and it deserves a short, repeatable routine of its own. Without it, the whole pipeline is open-loop: you have a process that should produce a good summary, but no mechanism that confirms a given run did.

The failures worth checking for are rarely ‘the summary is wrong in obvious ways.’ They are the quiet ones:

• Section coverage. Open the source’s table of contents alongside the summary. Every major section of the source should be locatable in the summary as a sub-heading or a clearly demarcated block. Missing sections are common when the source is long, when its chapter structure is weak, or when one section is far denser than the others.
• Hedge preservation. Pick three to five claims in the summary that read as confident assertions. Find the corresponding passages in the source. Models have a well-documented tendency to silently rewrite hedged statements (‘the author tentatively suggests…’, ‘X may be the case under conditions Y…’) as flat declaratives. If the source had qualifiers and the summary has dropped them, that is over-confidence drift — exactly the ‘vague accuracy’ failure mode the introduction warned about.
• Definition fidelity. For any specialist term the author introduces or uses in a stipulative sense, check that the summary’s gloss matches the author’s usage rather than the term’s general meaning. Over-smoothing of stipulative definitions is the single most common quality failure on theoretically loaded texts.
• Quote accuracy. Direct quotes should match the source word-for-word, including punctuation. Random spot-check two or three.
• Phantom references. Sample one or two of the ‘significant references’ the summary names — thinkers, texts, events. Confirm the author actually invoked them, in roughly the role the summary describes. Hallucinated references are rare in modern models but not zero, and they are catastrophic if they survive into downstream citation.

Budget roughly ten to fifteen minutes for this the first few times. After a handful of runs on the same kind of source you will learn which failure mode your specific genre is prone to, and the routine compresses to whichever subset still earns its time.

Visual Presentation

Pure Markdown text has several limitations: its formatting is relatively uniform, important points are difficult to emphasize visually, and the final appearance often depends heavily on the editor or rendering environment.

To improve readability and presentation quality, you can use the following prompt to transform Markdown into a richly styled standalone HTML document. This allows the content to be displayed in a cleaner, more structured, and more visually engaging way. For example, you can use the prompt below to convert your own content.

You are an expert editorial web designer, front-end developer, and long-form publication formatter.

Your task is to convert the provided markdown document into a polished, standalone HTML reading experience suitable for a premium think-tank, academic journal, or strategic analysis publication.

Requirements:

- Single self-contained HTML file
- Embedded CSS + JavaScript only
- Responsive layout
- Elegant academic/editorial aesthetic
- Serif body text, sans-serif headings
- Sticky left-side table of contents
- Reading progress bar
- Smooth scrolling
- Collapsible major sections
- Active TOC highlighting

Structure:

- Title area with metadata
- “Helicopter View” overview section
- Main numbered sections with concise subsection structure
- Minimal but clear visual hierarchy
- Subtle callout boxes for key critiques
- Simple vulnerability/risk badges
- Footer note

Style:

- Calm, intellectual, modern
- Long-form editorial publication aesthetic
- Avoid dashboard UI or generic markdown rendering
- Use restrained editorial colors and generous spacing

Content rules:

- Preserve the original structure and analytical tone
- Do not heavily rewrite content
- Do not over-summarize
- Reduce excessive bullet points where possible
- Prefer concise analytical prose with selective emphasis

Goal:  
Transform the markdown into a publication-grade analytical article rather than a raw markdown export.

After submitting the prompt, wait for the Agent to process the content. Once generation is complete, you will be able to preview the rendered HTML result directly.

Conclusion

The greatest value of AI is not to replace reading, but to help humans organise reading more efficiently. A high-quality academic summary is, in essence, a sequence of disciplined moves rather than a single generative act: first set the scope (what kind of summary, for whom, in what register), then fix the structure (the eight sub-protocols that govern coverage, order, faithfulness, and self-check), then generate, and finally validate against the source rather than against your own sense that the output ‘reads right.’

When AI is placed inside that four-beat rhythm — scope, structure, generate, validate — it stops being a text machine that returns whatever sounds plausible, and starts being a piece of research infrastructure: outputs you can trust enough to cite, hand to a collaborator, or build a literature note on top of.

Appendix

The sample input, the prompt, and the generated output (in both Markdown and HTML) are available in the Bandung Circuits repository open-sourced by Global South Insights.