Argentina has proposed the world's first legal framework for companies owned and operated entirely by artificial intelligence. On 4 June 2026, President Javier Milei used a Financial Times op-ed to invite AI firms to a deliberately unregulated, low-tax Argentina, and a bill already before the Senate would create a new corporate category, the "non-human corporation", carrying full legal personality with no requirement for any human shareholder. Two clarifications matter before the analysis: this is not legal personhood for the AI model itself, and it is not yet law.
What Argentina Actually Proposed
Milei's pitch rests on three stated pillars: a commitment to keep AI unregulated, the creation of the non-human corporation, and a low corporate tax rate. He framed the moment in historical terms, arguing Argentina should "become for AI what Amsterdam was for the age of sail."
The enabling legislation is a draft General Companies Law, authored by deregulation minister Federico Sturzenegger and sent to the Senate in late May 2026. It would replace Law 19,550, the corporate statute that has governed Argentine companies since 1972, making this the deepest reform to the country's company law in over five decades.
The bill creates two new structures, explained in detail by the Argentine fact-checking outlet Chequeado. A sociedad automatizada (automated company) is a firm whose business is carried out by autonomous algorithmic systems or AI agents, without requiring human employees or managers for ordinary operation. A DAO (decentralised autonomous organisation) is a blockchain-governed structure where holdings are tokens and decisions are executed by protocol. Both would receive full legal personality and limited liability, meaning they can own assets, sign contracts, and answer for damages with their own capital.
What It Is Not
For a topic this charged, precision is the story. Three corrections separate the reality from the headlines.
It is not personhood for the AI. The legal personality attaches to the corporate wrapper, not to the model running it. An AI agent gains capability only because the company around it can hold a bank account and sign a contract. The agent itself is not a legal person.
It is not regulation. Milei's first pillar is the absence of rules. The proposal is a deliberate non-regulation strategy, positioning Argentina as a single national special economic zone for AI, the approach Switzerland and Singapore once used to capture crypto and fintech.
It is not law. The bill stalled in the Senate in early June 2026, with the governing bloc lacking a safe vote margin amid an unrelated dispute over judicial nominations. Passage and timing remain unsettled.
The Collision With Anthropic and Harari
The proposal landed inside a single extraordinary week. On 4 June, the same day as Milei's op-ed, Anthropic published a warning that AI is approaching "recursive self-improvement" and argued it would "likely be a good thing" to have the option to slow or pause frontier development. Milei's message was the precise opposite: break free of restraint.
Five days later, on 9 June, the historian Yuval Noah Harari used his own Financial Times column to argue against exactly this kind of law. Granting legal personhood to AI-run corporations, Harari warned, hands them "a master key" to financial and political systems, and risks creating "not a company-state, but an AI-state." Milei responded on X the same day, thanking Harari and saying he was preparing a rebuttal.
The split is not fringe. On Milei's side, OpenAI's Sam Altman has publicly backed his vision of AI-driven growth. Against it, domestic critics including former lawmaker Elisa Carrió have called the plan "a catastrophic experiment for human dignity."
Why This Matters for Agentic Commerce
Strip away the politics and a structural point remains. The agentic commerce stack has been solving how an autonomous agent transacts. Argentina is the first jurisdiction to legislate who the agent legally is when it does.
The protocol layers are already live. The Agentic Commerce Protocol, from OpenAI and Stripe, handles checkout. Google's Agent Payments Protocol handles authorisation. The x402 protocol handles settlement. What none of them resolve is legal standing: when an agent owns inventory, signs a supply contract, or is sued over a transaction, which legal entity is on the other side. The non-human corporation is an answer to that question, and it is the first one any government has put into a bill.
For operators, the signal is not to incorporate an AI in Buenos Aires. It is that the legal infrastructure for autonomous economic actors has entered live legislative debate, years ahead of where most boards assume it sits. The jurisdictions that resolve agent legal identity will shape where agent-run commerce is domiciled, the same way company-formation rules already shape where firms register today.
What to Watch
Three markers will signal whether this is a turning point or a provocation. First, the Senate vote: whether Sturzenegger can assemble a majority, or the bill dies in committee. Second, imitation: whether any other jurisdiction proposes a comparable corporate category, which would mark the start of regulatory competition rather than a one-country experiment. Third, the infrastructure underneath it: Argentina has a letter of intent with OpenAI and Sur Energy for Stargate Argentina, a Patagonian data centre of up to 500MW and a reported $25 billion, the financial bet that the legal pitch is designed to attract.
Argentina has not changed the law yet. It has changed the question. The debate over who can own a company is now open, and autonomous agents are in it.
