Argentina Built the AI Law the US Couldn't Pass

Introduction

On October 10, 2025, the Argentine government announced that OpenAI had signed a $25 billion letter of intent to build a 500-megawatt data center in Patagonia. Sam Altman's words, posted on OpenAI's own site: "President Milei's vision for how AI can drive Argentina's growth and creativity is unmistakable and strong." That was three weeks before Argentina's midterm elections, the ones Donald Trump had publicly tied $20 billion in US financial support to Milei winning.

Eight months later, on May 29, 2026, Milei's government sent Congress a bill that would let a company be owned and run entirely by AI, with no human director, shareholder, or officer required by law. No country had ever tried it before.

What the Bill Actually Does

Argentina calls it the sociedad automatizada, the automated or non-human corporation. The draft amends the country's General Companies Law, Law 19.550, which has been in force since 1972 and is getting its first real overhaul in 54 years. The new category would grant a company operated by AI agents full legal personality and limited liability. It also recognizes DAO-style entities that record their operations on a blockchain. Human shareholders, in Milei's own framing, are "not required."

Milei laid out the logic in a Financial Times op-ed on June 4, co-authored with his Deregulation Minister, Federico Sturzenegger (summarized by the Buenos Aires Times, since the FT piece is paywalled). He compared the move to the 1602 founding of the Dutch East India Company and its invention of limited liability, kept AI regulation off the table entirely by calling premature rules "the deadly hand," and signed off with a line aimed straight at Silicon Valley: "We are open for business."

Here's the part that matters for everyone reading this, wherever you live. Limited liability for corporations works because a human is still standing behind it. When a company does something reckless enough, you can pierce the corporate veil and reach a person: a director, an officer, someone who can be sued or jailed. A corporation whose "CEO" is software removes that backstop. An Argentine AI specialist quoted in the local press called the combination of non-human corporations plus full deregulation "programmed impunity: human gains, social harm and responsibility shifted onto machines." That is the structure US tech companies have been trying to build at home and couldn't, which is where this whole thing is headed.

The Dinner Six Weeks Before the Bill

Peter Thiel was in Buenos Aires in April 2026. According to El País, citing Casa Rosada sources, he met Milei at the presidential palace at 2 p.m. on April 23. He had lunch with Milei's strategic advisor Santiago Caputo, went to a soccer match, and dined at the home of Federico Sturzenegger. That's the same Sturzenegger who co-wrote the non-human corporation op-ed and co-authored the bill. The dinner happened roughly 36 days before that bill landed in Congress.

It was Thiel's fourth meeting with Milei in two years. The New York Times reported that Thiel had bought a $12 million, 17,200-square-foot mansion in Buenos Aires and enrolled his kids in local schools. Milei's office denied offering him citizenship; two sources told the Times it was discussed. Thiel keeps more than 99% of his roughly $28 billion in assets in the US, and California has a 5% billionaire wealth tax on the November 2026 ballot, which would be about a $1.4 billion bill for him. A second passport and a residence in a country with no such tax saves him real money.

The congressional response is where it gets specific. Legislators asked the Foreign Ministry to disclose what was discussed in Thiel's government meetings, including whether a Palantir contract was on the table. Thiel co-founded Palantir, whose US government contract revenue ran $687 million in Q1 2026. The executive branch declared the meeting confidential. The question of whether Palantir was being negotiated has never been answered.

The timeline is suggestive, and the evidentiary limit is worth stating plainly. There is no leaked memo, no source on record, nothing showing Thiel wrote a word of this bill. The meeting was secret, so what was said is not in the public record. What is documented: four meetings in 24 months, a dinner at the bill co-author's home, a confidential designation slapped on it, and a congressional request that went nowhere. No single document connects that dinner to this bill. The circumstantial record is what it is — not more, not less.

The Minister Musk Copied

The other name worth holding onto is Sturzenegger, because he connects this to something closer to home. He is the minister whose deregulation ministry Elon Musk credits as the template for DOGE. Reuters reported that after Trump's 2024 win, Milei told Musk to contact Sturzenegger directly to learn how to dismantle a government. Sturzenegger confirmed Musk's team is in contact with his ministry. Musk called his "deep chainsaw" model "awesome" on X and described Argentina's approach as "a helpful model for the rest of the world."

So the person who served as the blueprint for the agency tearing through the US federal government is the same person who co-authored the world's first AI personhood bill. The same ministry that cut roughly 40,000 Argentine public jobs and shut down 250 national directorates, secretariats and undersecretariats, per El País and Reuters, is now redrawing what a corporation is allowed to be. The people who deregulated Argentina's state and the people building the US version run through the same man, Sturzenegger.

The philosophy isn't hidden, and it's worth sitting with for a second. Sturzenegger's whole project is removing the state from between private actors and outcomes. A non-human corporation is the cleanest version of that: a company with no human in the loop to regulate, tax the usual way, or hold responsible. The op-ed pairs it with a separate "Super RIGI" investment bill offering a 15% corporate tax rate and 30-year exemptions for projects over $1 billion, down from Argentina's standard rate of up to 35%. Opposition legislators called that one "a bespoke suit for Silicon Valley billionaires." Stack the two bills and you get the actual offer: bring your AI capital here, pay almost nothing, and structure the entity so no person is on the hook.

Who Benefits

Start with Thiel and Palantir. Palantir's business is processing government and corporate data at scale with minimal oversight. A legal framework where an AI agent can sign contracts, process citizen data, and run infrastructure as its own legal person removes the human liability layer that sits between Palantir-style systems and the people the systems affect. Argentine legislators thought a Palantir contract was close enough to ask about it in writing. The benefit here is a mix of money, market expansion, and the kind of cover that comes from a jurisdiction where the accountable party can be code.

OpenAI sits at the other end of the same play. It signed the $25 billion Patagonia letter of intent under Argentina's investment incentive regime, and at the same time it was lobbying at home for the opposite of accountability. In April 2026, OpenAI testified in support of Illinois SB 3444, a bill that would shield frontier AI developers from liability for "critical harms" (the bill's own definition covers events that kill or seriously injure 100 or more people, or cause $1 billion in property damage) as long as the developer filed safety reports and didn't act recklessly. OpenAI's March 2025 submission to the White House explicitly asked for "liability protections including preemption from state-based regulations." Argentina is the demonstration model. If AI-run entities can hold legal personhood and limited liability somewhere and function, that's a live example US companies can point to.

The bill's defenders have one real answer to this. Milei's op-ed says "ultimate beneficial owners must be disclosed," so a non-human corporation wouldn't be a fully anonymous vehicle, and somebody's name sits behind it on paper. The problem is that disclosure and accountability aren't the same standard. Knowing who profits from an AI-run company doesn't tell you who answers for it when the company causes harm, and the whole design of the structure is that no human officer is required to answer. You can know the owner's name and still have no one to sue.

Then there's Milei and Sturzenegger themselves. Milei's whole political story is that libertarian economics works, and he needs foreign capital landing to prove it. Thiel's mansion, OpenAI's letter of intent, the $20 billion currency swap from the US Treasury: every one reinforces the brand. That swap, per the Congressional Research Service, was the first use of the Exchange Stabilization Fund for a foreign government since Uruguay in 2002, and Trump explicitly conditioned continued support on Milei winning the midterms. Milei's party then outperformed its polling in November. The incentive to keep US tech figures visible and happy is structural, written into the financing.

They Couldn't Get It Here

In May 2025, US tech pushed a provision into Trump's budget bill that would have frozen state AI regulation for ten years. The Senate stripped it 99 to 1. The Illinois liability shield OpenAI backed is still contested. Across more than 40 states, AI accountability bills keep getting introduced, and the industry's lobbying arm has had to fight them one statehouse at a time. The domestic route to zero AI accountability is slow and public, and so far it's losing.

The Argentine route is one minister and one president signing off on one bill. That's the actual story underneath the data center announcements and the mansion purchase. You don't need 50 state legislatures if you have personal access to the two men writing the law in a country you're already moving capital into. The framework US tech couldn't pass at home is now sitting in a Senate committee in Buenos Aires.

It isn't law yet, and that caveat is real. Milei lacks a reliable majority, and Argentina has a long record of ambitious bills that stall or get reversed. Critics inside the country are loud: opposition figure Elisa Carrió called the proposal a "catastrophic experiment for human dignity." And the man who wrote Sapiens, Yuval Noah Harari, published an FT rebuttal on June 8 (covered by the Buenos Aires Times) arguing that granting legal personhood to AI agents hands them "a master key" to financial and political systems, and warning that a country could become "not a company-state, but an AI-state." Microsoft's AI chief Mustafa Suleyman agreed, citing the argument that AI agents should have "no more rights or freedoms than a laptop."

The Bottom Line

The bill that lets an AI own a company with no human accountable for the damage is the same structure US tech lobbied for and lost on at home. Same idea, different venue. A framework that needs 50 state wins in the US needs one dinner and one minister in Argentina, and the billionaire with the most to gain from it had that dinner six weeks before the bill appeared.

The open question is whether this is a one-country experiment or a template. The EU withdrew its AI Liability Directive in October 2025; even the most cautious major jurisdiction backed away from holding autonomous systems accountable. Argentina is filling that vacuum from the other direction — and doing it with a network of US tech capital that already knows how the domestic version of this fight ends.