Argentina Wants AI to Run Companies With No Human Owner

If you have been following any of the AI news, one of the most interesting things moving right now is happening in South America. A government down there wants to take artificial intelligence and make it able to form a corporation, where there is no actual human being in charge of the so-called company. The country is Argentina. And I want to walk through it honestly, because the headline version is scarier and more finished than the real version, and you deserve the real one.

What Is Actually True, and What Is Not

Here is the part to get right before you repeat it anywhere. This is a proposed bill, not a law. In June 2026, Argentina's president Javier Milei and his deregulation minister Federico Sturzenegger laid out a plan, including an op-ed in the Financial Times on June 4, to create a brand new legal category they call the non-human corporation. A company that can be operated by AI agents, where human shareholders may participate but are not required. As of the middle of June it has been sent to Congress and it still faces debate. Nobody has signed artificial intelligence into a boardroom yet.

One more distinction that matters. The bill does not make the AI itself a legal person. It makes legal room for a corporation that is run by AI. And that is the clever, slightly unnerving part. Corporations are difficult to sue. I am not a lawyer and I do not play one on TV, but the corporate shell is the ultimate liability shield. Now picture taking an AI, dropping it inside that shell, and giving it the ability to function on its own. The shield is real. The human standing behind the shield may not be.

Why Argentina, and Why Now

Easy to assume the country got fooled. I do not think that is it. Milei is a self-described anarcho-capitalist who was selling this worldview long before any Silicon Valley money showed up. His pitch is to make Argentina the number one destination on earth for tech and AI capital escaping regulation everywhere else. His own line was that he wants Buenos Aires to become for AI what Amsterdam was for the age of sail. Argentina is not the mark here. Argentina is the volunteer.

That is a sharper and more useful way to see it. A whole country is offering itself up as the sandbox for an idea that, until now, people could only test at the scale of a single private city. Why build one experimental town when a national government will hand you the entire legal system to try it in?

What Is Peter Thiel Doing in All of This

You will hear that Peter Thiel is running this. Be careful with that claim, because the honest version is more interesting and harder to knock down. What is confirmed: Thiel relocated to Buenos Aires, bought a home reported around 12 million dollars, and met Milei at the Casa Rosada in April 2026. What is not confirmed is any signed role in writing this bill. His connection is the philosophy and the pattern, not a paper trail.

And the pattern is the tell. Back in 2009 Thiel wrote, in plain words, that he no longer believed freedom and democracy were compatible. He has spent the years since funding ways to route around ordinary democratic government. He backed the Seasteading Institute and its floating startup-countries. He funds the venture firm behind Prospera, the private charter city in Honduras. He co-founded Palantir, the surveillance-software company. So when a libertarian president starts building a deregulated AI zone and the architect of exit-from-democracy buys a house down the street, you do not need a signed contract to notice the shape of it. Just do not overstate it on camera. Say he funds the fund that funds these experiments, because that part is true and it holds up.

What Did Somebody Sell Argentina, and How Hard Was It

This is the question I kept circling, and the answer flips the whole story. Nobody sold Argentina a piece of software to run the country. The only real transaction on the public record is a billionaire buying real estate. What got sold was an idea. Deregulate AI completely, give the machine a corporate shell with limited liability, become the Amsterdam of artificial intelligence. And here is the kicker on how hard the sale was. It was not hard at all, because the customer wrote the sales copy. Milei published the op-ed himself. When the buyer is doing the pitching, you are not closing a deal. You are watching someone who already decided.

The Question Nobody Will Answer. Who Do You Sue?

This is where it stops being a thought experiment and becomes your problem. In normal life, if a company sells you a product and it hurts you, or you get damaged in the regular execution of your day, there is somebody to make a complaint to. Whether anything happens depends on attorneys and how hard you both want to fight. But there is a who. There is a name on the other side of the table.

Now take that human out. Hand the corporation to an autonomous system with its own level of capability, and let it function on its own. Who do you sue when it goes wrong? Who gets held accountable when the entity that harmed you has no person attached to it?

We are not guessing here, because there is a live preview, and it runs the opposite direction from what you would expect. In Honduras, a private charter city called Prospera is suing the country of Honduras itself, through international arbitration, for a quantified floor of 1.6 billion dollars and a ceiling north of 10 billion. That is roughly a third of the entire national budget. Honduras voted, unanimously, to repeal the law that created these private zones, calling it a violation of national sovereignty. Its Supreme Court ruled the zones unconstitutional. And the country still faces the bill. Sovereignty on paper, invoice in practice.

That is the accountability arrow flipping. When governance gets privatized, the corporation can sue the nation, and the citizens are left in private courts they never voted for and cannot vote out. America already wrestled with a smaller version of this in 1946, when a company that owned an entire town tried to enforce its own rules on its own streets, and the Supreme Court said no, the Constitution still applies inside your fences. We answered part of this question almost eighty years ago. The new wrinkle is that this time the thing inside the fence might be a machine. You cannot vote out a model. You cannot subpoena a weight.

Then What Happens Next

That is the part that keeps my attention. It rarely stops at the first step. Today it is the ability to form a corporation. Tomorrow the question becomes whether the autonomous entity gets more room, more reach, more ability to do its own bidding. Does it start to get power that touches the government itself? There is a separate Argentine program, a so-called social digital twin meant to simulate policy using citizen data, and the government has flatly denied that Palantir or any private company is building it. Worth watching, but keep it labeled. Right now it is a denial and a rumor, not a confirmed deal.

Everything about this lands in one of two buckets, and there is very little in the middle. It is either going to be really, really good, or really, really bad. It could pull genuine investment and jobs into a country that needs them. Or it becomes what one Argentine critic already called a catastrophic experiment for human dignity. I do not get to tell you which, and neither does anybody selling you certainty.

The Honest Take From One Human to Another

I will tell you the same thing I told you when the government yanked Anthropic's Fable 5 model off the shelf three days after it launched. Use these tools. They are genuinely beyond what most people realize, and I build on them every single day. But do not confuse a tool you rent for a thing you own, and do not cheer for systems that quietly remove the human you would need to hold accountable when something breaks. It is the same thread I keep pulling when I ask who the AI actually works for, and what happens to regular people when the machine replaces the workers but still needs someone left to sell to.

Before Argentina becomes the next experiment, before any country hands a corporation to a machine, somebody has to answer the one question under all of it. When the company runs itself and the AI makes the call, who do you sue? Until there is a real answer, the rest is just a very good soap opera with a very expensive cast.