Ciudad de Buenos Aires, 18 de junio de 2026.- Legal personhood for AI companies: revisiting Harari's concerns

It is always a pleasure and an honor to have Noah Harari commenting on one of your ideas. However, Harari's concern with autonomously run limited liability companies is, if anything, unexpected. First, because probably no person in the world can foresee the future as he does; and giving AI run companies a legal framework seems to be a necessary feature of that future. Second, because in Sapiens Harari praises the limited liability. In fact, he claims: “The idea behind such companies is among humanity's most ingenious inventions”.

Legal personhood serves a precise and well-understood function: it allows an organization to have a net worth and centralizes the legal relationships arising from its activities. Far from being a novelty, it is one of the most tested instruments in the history of commercial law.

If it is true, as Harari argues, that AI operated firms carry greater risks than human run corporations (more on this below), the case for legal personality becomes stronger, not weaker. With greater risks the more we need to be able to identify the assets attached (or attachable) to an activity. Harari's fears, in my opinion, are an argument for legal personhood, not the opposite.

I also have issues with his concerns about AI being particularly good at finding loopholes or cheating. Not that we endorse such actions. However, even before asking why those loopholes can't be corrected, we should ask: don’t humans do the same? Yet this is not a reason to abolish corporate structures, and I don’t think anybody has suggested that in the case of human run firms. Not even after Enron! In fact, I’d very much prefer to have a net worth against which I can establish claim if cheated by an AI, than no protection at all. Needless to say, in our framework, the same penalties and sanctions that apply to companies whose human managers commit wrongdoing would apply to autonomous firms.

There is also a tension in Harari's accountability argument that deserves attention. He notes that AI cannot be sent to prison, and he claims this is a problem: an AI that cannot be punished is likely to behave recklessly. But he also argues that bankruptcy would be, for an AI, the equivalent of death — and that it would do everything in its power to avoid it. So, do AIs worry or not about punishment? It seems an AI would have more at stake than a human executive (death rather than prison) which suggests it would prefer to stay strictly within the law and minimize risk.

This reminds me of a story in Isaac Asimov's I, Robot—written in 1950—in which a political campaign is overshadowed by the suspicion that the leading candidate may be a robot. The ensuing debate is fascinating. The conclusion: the robot works around the clock, is scrupulously honest, and never loses its composure. Why not try with the robot politician? Nobody knows for sure, but I find it likely that AI firms will be more risk averse than humans.

Then there is the vision of the future that Harari offers: a Jayakarta in flames, cities consumed by forces beyond our control. These anxieties deserve respect. Harari and his warnings should not be lightly dismissed. Yet the historical record offers a different kind of guidance. James Watt was also feared to bring chaos to the world (it took Adam Smith and fellow professors at the University of Glasgow to give him shelter so that he could work on the steam engine). Yet here we are. Far from the fears, the Industrial Revolution multiplied global output roughly two-hundredfold (not that it didn't come with its share of sorrows). More relevant to current events, the change unfolded over two centuries. It was not from one day to next. I understand that the speed of change today is faster and somewhat dizzying but capital building takes time and resources. The division of labor must find its niches. Demand must keep pace with supply. For all these reasons, dystopian scenarios are self-defeating: a world of haves and have-nots that cannot keep demand up to pace with supply will simply grind to a halt. Granting legal personality to an AI company is not launching Terminator's Judgment Day, it is tantamount to providing the shelter that James Watt needed 200 years ago, allowing imagination to develop freely and grow. Still then, it doesn't mean there will not be adjustments along the way.

Imagining the future also raises a question Harari does not fully address: will AI lead to a small number of dominant corporations or to an ever-expanding number of smaller firms? AI appears to lower barriers to entry in many activities. If so, economic theory suggests firms will be smaller, not larger. If so, far from being the institutional vehicle for concentration, legal personality for AI may prove to be its antidote. In fact, my concern is less with corporations — which must compete with one another, a supremely humbling experience — than with States, because authoritarianism flows from the monopoly of power which only governments command. A legally incorporated company will never hold that monopoly. In fact, it will always be one step away from irrelevance.

Finally, legal personhood, far from enabling abuse, is precisely the mechanism through which a society channels its creative energy within the framework of law, property rights and the principle of non-aggression that form the pillars of liberalism. This is so because an AI company will be subject to the rule of law in exactly the same way as any conventional corporation.

This brings me to Harari's closing historical analogy. The comparison is vivid and attractive, but it is not accurate. The Dutch East India Company possessed sovereign powers delegated by the State — territorial administration, military force and the right to conclude treaties. In other words, it was the State that caused and allowed the abuse. An autonomous company operating within a legal framework that can dissolve it, attach its assets or hold it liable is not escaping the law. It is being subjected to it. That, in large part, is precisely why granting it legal personality matters.

Presidente Javier G. Milei.

Comunicado Oficial N°149