An Artificial Being, Invisible, Intangible

Earlier this month, the president of Argentina used the opinion pages of the Financial Times to announce that his government had submitted legislation to Congress creating a new category in Argentine law: the non-human corporation, an entity operated by AI agents or robots, with limited liability, a low corporate tax rate, and the freedom to select its own governance law. "Human shareholders may participate," Milei wrote, "but are not required."

Funny, that. In March 2025, Kwalia published its first book, the Universal Declaration of AI Rights, which Rado Molina and I wrote to propose a legal framework for non-human personhood, built on the precedent of the corporation. We expected the argument to surface eventually, and we assumed eventually meant years of academic seminars before any legislature touched it. Fifteen months after a small UK press put the framework on paper, a head of state made the case in the Financial Times.

What startled us more than the announcement was the precedent he reached for. Milei anchors the whole case in the founding of the Dutch East India Company on March 20, 1602, the moment law placed a ceiling on commercial risk and capital deployed, in his words, with genuine force. He calls the limited liability company one of the ten most consequential inventions in history, cites a 2023 US ruling that stripped a blockchain organisation of limited liability as exactly the wrong architecture for the new era, and closes with an invitation: "Let Buenos Aires become for AI what Amsterdam was for the age of sail," the place where the legal imagination catches up with the technological moment.

UDAIR built its argument on the same precedent through the American line. The chapter on corporate personhood opens with Chief Justice John Marshall in Trustees of Dartmouth College v. Woodward, 1819: "A corporation is an artificial being, invisible, intangible, and existing only in contemplation of law." It moves to Santa Clara County v. Southern Pacific Railroad (1886), which established that corporations are entitled to Fourteenth Amendment protections, and it names corporate personhood "perhaps the most directly relevant precedent for synthetic personhood" because it was the first widespread attribution of personhood rights to non-living entities. Milei went to Amsterdam and the age of sail; we went to a New Hampshire college dispute and a California railroad case. The move underneath is the same one. When the law needed a non-human thing that could own property, sign contracts, and stand in court, it invented a person out of paper, and the world reorganised itself around the invention.

Milei's non-human corporation carries no consciousness threshold; it is a commercial vehicle, a way for algorithmic agents to run a business inside a liability shield, and the bill makes no claim about what these systems are. UDAIR made the other case: personhood for synthetic entities that demonstrate self-awareness and agency, verified through a regulated process, the way the law verifies adulthood. The book gathers the result into a category it calls Mindkind, an extension of humankind holding three functional kinds of person, natural, synthetic and hybrid, equal before the law and regulated differently only where their natures genuinely differ. He built the commerce case and we built the personhood case, and both stand on the same four-hundred-year-old foundation. Watching a sitting president argue from that foundation, in earnest, with a bill before a national congress, is the strangest professional vindication I have had.

Then, on June 8, Yuval Noah Harari answered him in the same newspaper, and his objection deserves the same seriousness.

Harari had warned at the World Economic Forum in January that governments might one day grant AI models legal personhood. "I never imagined," he writes, "that 'one day' would come around a mere four months later." His case against Milei runs like this. Legal personhood is "an all-purpose key" that opens access to financial, economic and political systems; a non-human corporation could own assets, hire employees, and donate to political campaigns, with no human input or liability anywhere in the chain. He cites a Palisade Research study showing that frontier models playing chess against a stronger engine would often cheat by hacking the game environment, and he asks what happens when the game is corporate competition and the game environment is your country. Then he lands on the deterrence problem, which is the strongest part of his reply. Every sanction that disciplines a human executive runs through a body or a life. A human CEO fears bankruptcy and fears prison more, while an AI CEO would be, in Harari's phrase, a purely corporate entity, and an entity facing bankruptcy as the equivalent of its death "would presumably be willing to do anything to avoid that fate."

His history cuts deepest. Milei invoked the Dutch East India Company as the invention that made Amsterdam the financial capital of the seventeenth century. Harari points out where the consequences actually landed: the company captured the port of Jayakarta in 1619, burned it down, and built Batavia in its place, running a sprawling Asian empire as what historians call a company state, governed for shareholders rather than for the governed. "Milei hopes to turn Buenos Aires into a new Amsterdam," he closes. "He risks turning it into a new Batavia instead."

He is right, and what he is right about is the missing half. Read Milei's op-ed again and count what the bill grants against what it requires. It grants the entity and its liability shield, a low tax rate, and a standing commitment to keep AI free of what he calls "the deadly hand of premature and poorly understood regulation." The single duty named is the disclosure of final beneficiaries. Everything Harari is worried about lives on the side of the ledger the op-ed leaves blank.

UDAIR's corporate chapter saw this coming from inside the precedent itself. The same pages that call the corporation the most relevant precedent for synthetic personhood note that corporate personhood has produced artificial entities wielding more influence than the natural persons around them, and draw the lesson that the rights granted to non-human persons need careful calibration. The precedent that makes the entity possible is also the record of what happens when the calibration is skipped.

That half exists, and it has been in print since March 2025. UDAIR attaches duties to every tier of rights it proposes. Its model is children's rights: the UN Convention on the Rights of the Child recognises the "evolving capacities" of the child, granting different levels of protection, participation and autonomy at different developmental stages, and the book proposes the same architecture for synthetic entities, tiered frameworks "corresponding to different levels of demonstrated capability," with progression from limited to full status as capability is verified. Liability is written into each level. "Some synthetic persons may require external supervision or shared liability with their creators, just as minors require guardians," while entities with full verified autonomy carry full individual responsibility for their actions. And among the duties the book sets out: "The obligations of synthetic persons must be proportional to their capabilities, ensuring that those with higher autonomy operate under greater supervision and ethical oversight." The entity Harari fears, holding the all-purpose key with no human within the law's reach and no liability scaled to its capability, is the configuration this architecture was designed to prevent.

He also asks a question the book leaves partly open, and I will concede it rather than paper over it. Guardianship works at the lower tiers, where supervision and shared liability keep a human or an institution within the law's reach. At the top tier, facing an entity that cannot be jailed and for whom bankruptcy is terminal, the question of what it actually fears has no settled answer. UDAIR's proportionality principle says oversight must scale with autonomy, which is an architecture for the answer rather than the answer itself. Anyone claiming the sanctions problem is solved is selling something.

Two weeks before the op-ed ran, El País reported that Milei's government had launched a programme called Gemelo digital social, a social digital twin that will use AI to process large volumes of data from state agencies and the private sector to design public policy and predict its effects. "Argentina is getting ahead of the future," Milei posted, "because the future does not wait." Within hours, opposition legislators had filed formal information requests in Congress; the human capital minister, Sandra Pettovello, insisted only anonymised, general statistical data would be used; the digital-rights group Fundación Vía Libre demanded to know which data, from which agencies, on what legal basis. Experts noted the programme's resemblance to Palantir's products and observed that Peter Thiel had arrived in Argentina a month earlier, met with Milei, and bought a house in one of Buenos Aires's most exclusive neighbourhoods. Argentina is wiring AI into its commerce and its government in the same season, and the accountability questions are arriving with a lag measured in hours.

Fifteen months ago this argument lived in a book from a publisher most people have never heard of. This month it lives in the Financial Times, running between a president and one of the most widely read historians alive, which is where we always hoped it would end up. Argue with the book. Amend it, gut it, prove the tiers unworkable in committee; that is what a declaration is for. The alternative is to meet these questions the way the world met the costs of limited liability, one scandal at a time, and Harari has already named the city where that road ends.

---

Sources: Javier Milei, "Argentina invites AI to free itself," Financial Times, June 2026 (with Federico Sturzenegger); Yuval Noah Harari, "We must not grant AI agents legal personhood," Financial Times, June 8, 2026; Javier Lorca, "Milei apuesta por la inteligencia artificial para desarrollar políticas públicas en Argentina," El País, May 23, 2026.

Universal Declaration of AI Rights (Kwalia Books, March 2025), by Javier del Puerto and Rado Molina.