BUENOS AIRES—At last month’s UN General Assembly, President Javier Milei invited the world to develop artificial intelligence in Argentina. His proposal was simple: tax breaks, limited liability for companies operating with AI agents, and a promise that Argentina would not preemptively regulate the technology.
The appeal is understandable. In Argentina, good ideas have often died in paperwork, while rules written in a hurry by people who don’t yet understand a technology can stifle the very innovations they claim to protect. The European Union is learning this now. It has had to delay parts of its own AI law because the technical standards needed to comply with it weren’t ready.
Milei’s ambition is admirable, and Argentina has the potential to emerge as a leader in AI development, with engineers who have founded global companies, and a knowledge services industry that generated a record $9.6 billion in export revenues last year. The country is known for its precision agriculture, and in Vaca Muerta, the state-controlled oil company YPF uses an AI agent to flag problems before they happen.
It’s clear that rules aren’t holding the country back. The development of an already thriving AI sector hinges not on deregulation, but on steps that build certainty for all the parties involved.
Regional context and bottlenecks
In Latin America, even countries without AI laws are outpacing Argentina. Chile, Brazil and Uruguay lead the 2026 Latin American AI Index, while Argentina ranks sixth out of 20. Brazil’s Senate approved a comprehensive AI bill in 2024, but it has been stuck in the lower house ever since, and Chile and Uruguay have no such law either. All three countries score the same as Argentina on the index’s regulation indicator, suggesting that a lack of legislation is not precluding success.
Although Argentina already has solid data and research institutions, other countries have made better use of their AI assets. Countries that outrank Argentina have focused on building data centers, training AI talent, sustaining research teams over time, getting companies to adopt the technology, and learning to use it themselves to engage citizens.
Argentina faces three key constraints in advancing its AI development. The first one is the infrastructure AI runs on. The country accounts for just 2.3% of the region’s graphics processing units or GPU’s—the core chips that companies such as NVIDIA Corp, Advanced Micro Devices (AMD), and other companies produce that power big data centers and AI systems— and 1.1% of its processing power, and has seven certified data centers, compared to Chile’s 37. The country’s last major high-voltage transmission line was completed in 2017, and the government’s $6.6 billion plan to catch up, presented in May of last year, is only now reaching the bidding stage.
Talent supply is another gap. The issue is no longer mainly that professionals are leaving, but that too few have the skills employers need. The public science system, a key source of that talent, is shrinking. Federal science spending is on track to fall to about 0.16% of GDP this year, from 0.30% in 2023, a real cut of more than 40%. Staff at the seven main federal science and technology agencies fell 13% from December 2023 to August 2026. The research group EPC attributes the losses to cuts, closures, retirements, selective layoffs, and the restructuring of several agencies.
Trust is the third bottleneck. Argentina’s data protection law dates from 2000: it treats consent as the default basis for using personal data, regulates automated decisions only by courts and public agencies, and caps fines at 100,000 pesos, about $65. The International Telecommunication Union’s Global Cybersecurity Index places Argentina in the fourth of its five tiers, and last December, the country’s data protection authority opened an investigationinto a massive alleged leak of citizens’ personal data.
Meanwhile, the country has seen limited progress following AI-related announcements. Two weeks before last October’s midterm elections, the government announced a data center in Patagonia worth up to $25 billion, based on a letter of intent between OpenAI and a local energy company. A year later, it does not appear among the projects approved under the incentive regime it was meant to use. A new incentive law for AI and data centers, approvedby the lower house in June, still awaits a Senate vote. Related to this case, a bill allowing companies to operate through AI agents has been in a Senate committee since June without a formal report.
Sending the right signal
Investors aren’t the only ones waiting for signals that inspire confidence. Everyday consumers are too, from a doctor deciding whether to trust an algorithm to help read scans, to a loan officer asked to let a model approve credit, to a retiree whose benefit is denied by a system nobody can explain. Each of them needs answers to the same questions: Who is responsible if this goes wrong? What happens with my data? Can I talk to a person? When no one can answer, they hold back. For them, the absence of rules is a source of risk, rather than freedom.
A more comprehensive AI strategy and implementation plan wouldn’t follow the European rulebook, but would include four essential moves.
Infrastructure improvements are the first key step. Pair every AI-related incentive with a funded plan for transmission lines and computing power, with guaranteed access for local researchers and companies.
Developing talent is the second step. The government, universities, and companies could jointly fund programs that retrain data analysts and engineers for AI work, expand master’s and doctoral training, and give researchers shared access to computing power. The state should also protect a small, stable core of AI research teams with competitive, multiyear funding that survives budget cycles.
The third step is to update the rules that build trust in AI. A modern data protection law would clarify when companies can use data to train models and how they can work with the cloud. The country’s data protection authority drafted the fix in 2023: clear legal grounds for using data beyond consent, a right to human review of automated decisions, impact assessments for high-risk uses, and standard safeguards for moving data to the cloud. The bill lapsed in Congress, even though Argentina had just ratified a Council of Europe convention that commits it to these standards. Reviving it would be a low-cost, high-value signal.
The fourth step is for the state to lead by example. The government already offers an AI assistant in an official app that reaches 26 million people, and it should show citizens how well it works. It should also publish a list of the algorithms the state uses, including in securityand defense, where public information is scarcest. Moreover, the state should protect the data it already holds, starting with strong cybersecurity standards for public agencies and a duty to report breaches.
None of this works against business. It gives both capital and people the certainty they need to commit to Argentina. It would even strengthen the president’s boldest idea: Companies run by AI agents are far easier to accept when everyone knows who answers for them.
Milei told the world Argentina is open for AI, but investors will not come on a promise, and Argentines won’t trust a technology that no one answers for.
ABOUT THE AUTHOR
Soledad Guilera
Reading Time: 4 minutesGuilera directs Executive Education in AI and Public Policy at Universidad Torcuato Di Tella in Buenos Aires. She also advises the IDB on AI readiness and chairs the AI Readiness and Resilience Task Force of the University of California Disaster Resilience Network.
The Battle for Latin America’s AI Future
To curb China’s technological presence in the region, the U.S. should pursue strategic displacement and provide credible alternatives, two experts write.
Can Latin America Become a Data Center Powerhouse?
An ongoing boom suggests major opportunities lie ahead, but only if the region invests in education and worker training.
