Yuval Harari called for a global ban on AI legal personhood at Davos.

Ok. But what's the alternative?

<iframe width=”560” height=”315” src=”

Here’s that Harari interview at Davos everyone’s talking about. The discussion of legal personality for AI starts at minute 26.

Imagine my surprise that famous historian and public intellectual Yuval Harari has a very strong opinion on the previously obscure topic of AI legal personality. His opinion? Just don’t. As an expert on the identification of persons under the law, I’ve been taken aback at how rapidly non-lawyers have suddenly developed very strong opinions on this fascinating, but esoteric, legal problem. (I’m sure Lawrence Solum didn’t realize that he was establishing his legacy when he wrote his seminal article on the question back in 1992s. He probably thought it was a fun, Sunday afternoon, zany idea.)

Harari does a great job summarizing the problem with AI legal personhood. AI isn’t a corporation, so it’s a bad fit for legal personality, as we here at Future of Citizenship have argued before (though it would have some benefits and doesn’t deserve to be dismissed out of hand without a thorough analysis of the pros and cons.) Legal personality is a fiction granted to corporations that are run by humans, and not really to the objects themselves. Yet, despite being a fiction that can be taken away by courts at any time, some US states have already moved to ban legal personhood for AI, and now, Harari is calling for a global ban. (Never mind that global bans are hard - that’s a problem for another day.)

Why? Because they correctly intuit that legal personality for AI would open the door to non-fictional person status for AI - aka a legal identity for AI. (You’re a non-fictional person, not a fictional device created by the courts.) And if you want to see how great the conversation around AI rights and identity is going, you can hear the fear in this interview with Yoshua Bengio, or watch the revulsion on Tegmark’s face in the above video.

But for the moment, I’d like to turn back from the explosive question of robot rights, and return to a discussion of AI legal (corporate) personality.

The Responsibility Gap Versus the Liability Trap

Legal personality was not invented to imbue objects and concepts with magical properties they don’t actually have. European society in the middle ages had such a system, that of deodand. If a bale of hay fell on your head, the hay could be guilty.

We no longer have such a system. However, we do sometimes treat groups of humans as a single, fictional “person,” for the purposes of distributing liability and risk across the entire group, rather than concentrating it within a single decision maker. That allows leaders of groups to make decisions without a crushing level of personal risk that would paralyze them. Legal personality allows modern society to work. As the president of my condo board, I can install a new staircase in our building, safe in the knowledge that I will not be personally liable most of the time. You benefit from the fiction of legal personality every single day.

The problem with AI is different from the problem legal personality was designed to solve. AI is not a group of humans with a leader, it is a non-human, semi-autonomous intelligence. With time, it will become both more autonomous and more intelligent. Yet, our legal system is only equipped to deal with humans that are autonomous and intelligent. We no longer treat bales of hay as deodands, capable of actions on their own. If a bale of hay falls on someone’s head, we blame the owner of the bale, or the owner of the barn, or some other human, not the hay. If a corporation owns the barn, we blame the corporation, which is another way of saying, the group of humans as a whole. But AI is not a group of humans. It is an intelligent, autonomous bale of hay - a real deodand.

Yet, the fact that our legal system doesn’t have a ready solution to the problem of AI autonomy doesn’t make the problem go away. In fact, Harari himself spends most of the early part of the above interview outlining many of the ways in which AI is set to become even more independent and agentic. But he doesn’t propose a solution. And before we man the barricades in our Les Miz tee shirts, calling for a global ban on legal personhood for AI, we need to be ready to propose an alternative.