Katherine Forrest: Right. So, personhood is where all of those threads really come together. I wrote actually two essays about legal personhood for AI, not advocating one way or the other, but just sort of talking about the conceptual frameworks that apply to non-human entities over time and actually even humans over time and differences between how we've treated different kinds of people or categories of people at different points in time. So I wrote one of those articles for the Yale Law Journal Forum and also one for the Fordham Law Journal. And in both of those I talked about legal, you know, personhood as not fixed or biological, but really as something that is flexible and as I and others have called it mutable. It can change. You can have a a corporation that can start off with certain rights and you can add rights to it, but you can also, and we have over the history of this country, started off with differences between human beings and the amount of legal personhood and the rights as legal people that certain groups of humans, let's just take women as an example, have had. So judges and legislatures for AI are going to be presented with questions that will require them to weigh the harms of granting or denying legal personhood to AI, agency and responsibility of AI systems. And as we then, and this is the intersection with sentience, as AI systems may obtain more awareness, or if they do, or if we're wondering if they have, there may be some moral questions and personhood questions that get attached to that. So there's that separateness, but also that overlapping piece that we talked about.
Scott Caravello: And that seems like a perfect segue into talking about the specific laws and proposals that have been put forward in the US. And so I guess I'd categorize the three that have been successful into what are called the bright line anti-personhood bucket. They are broad but clear prohibitions on personhood. So Idaho, for example, enacted a law in 2022, actually. Which provided that not only AI but also environmental elements, non-human animals and inanimate objects, won't be granted personhood in the state. And then Utah followed two years later, prohibiting governmental entities from granting or recognizing personhood for AI and other non-human categories. And then Tennessee, just a few months ago, joined the camp too by amending its general statutory definition of person to exclude AI, a computer algorithm, software, computer hardware, or any type of machine.
Katherine Forrest: Yeah, and let me just pause for a moment on the aspects of those state you know, laws and the various bills that are winding their ways right now through other state legislatures to talk about the non-human animals or inanimate objects, because that may not strike people as necessarily intuitive in terms of why. For some of it, it's not just about AI. There have been instances both in this country as well as for instance New Zealand and other countries where a mountain, a river, have been granted a kind of legal personhood. The reason for that is, or the concept behind it, is to grant legal personhood to say a a river or a mountain and thereby enable or allow that entity or somebody on its behalf to sue in order to create certain protections for the mountain, for the you know, or for the river. And so that's part of what's behind that. So here we've got that precedent, but also AI as an inanimate object. And also a non-human. So, all of these things go together. And the successful laws that you've just talked about, Scott, share a few features. And they're short, they're to the point, and you know, thou shalt not marry an AI. Idaho and Utah also don't focus only on AI, but they also deny personhood, as we've said, to these other non-human objects including things like and this is a new one for me, atmospheric gases. Like why? I mean, I didn't read the legislative history on this, but… what are they talking about?
Scott Caravello: I mean my best guess is something having to do with like oxygen and carbon and then getting at your point about suing and being sued over climate change issues. That's… that's as best as I can think of.
Katherine Forrest: I mean, these are perhaps defensive maneuvers. And Tennessee is framed a little bit differently from those, but you know, as you'd mentioned, it amends the general statutory definition of personhood. And the goal though is the same. It's to keep personhood limited to humans and recognized legal entities because remember, corporations are legal persons. I mean they are considered to be a person under the law and a corporation can own property, it can hire and fire, it can, you know, have employees, it can have debts, it can have obligations. So, you've got this fiction of a legal person in the form of a corporate entity. So, there'll have to be, of course, carve-outs for all of that.
Scott Caravello: Yeah, and then, actually, I think what's interesting about the Idaho law in particular is that it specifically makes clear that what it's doing is not impacting corporate personhood. The law specifically says it doesn't revoke the status of municipalities, organizations, corporations, or other legal or business entities recognized before July 1st, 2022, which is a reference to when the law came into effect.
Katherine Forrest: It’s interesting because you could conceptually, actually incorporate an entity, the only asset of which was an AI model. And you could have a corporation whose sole asset was and sole mode of acting was through an AI model. And if you had a corporation then that could obtain legal personhood, you might, by a derivation, enable AI to act through that corporate entity and obtain legal personhood. So, it's an interesting, you know, set of carve-outs, but I'm not sure that they're, you know, if your goal and it's somebody else's goal, it's these other states' goals, if their goal is to achieve that, I don't know if it's ironclad, but we pause on that point about corporations again, because, you know, AI, unlike corporations, raises different problems. And this is now again where that overlap occurs that I talked about at the beginning. You know, in a corporation you're not talking about necessarily the morality of its own innate desires to act or not act. It's acting through the individuals within the corporation, whether it be the board, the shareholders, the management. But AI can have, potentially, different cognitive abilities, over time, that might give it a set of desires to take action in a particular way that is different from a corporation and that would lead us to, you know, a real fork in the road between the two.
Scott Caravello: And I think that leads us perfectly into the other bucket of these state proposals, all of which have so far been unsuccessful or are still pending in their respective state houses. But they get a bit more specific about what they're aiming to prohibit and even get at some of the concepts that you were just discussing, Katherine. And I think I'd pick out Ohio's bill, HB469, as a really good example of that. And again, that one is seemingly dead. It's been stalled in committee for quite some time. But not only would it deny AI systems personhood, but it would also declare that AI will not be considered to possess consciousness, self-awareness, or similar traits of living beings. And it includes a ban on AI entering into a marriage or a domestic partnership, whether with human or with another AI. And, you know, if I had to take a guess, I think what you'll say is really interesting, Katherine, is that it's declaring that AI will not be considered to possess these traits of consciousness and sentience, which is, of course, irrelevant to the question of whether the technology develops to that point and whether they actually do. So, it's staking out a line in the sand regardless of how the technology develops and what traits they come to possess and, you know, how we as a society will then have to grapple with that.
Katherine Forrest: You know, it reminds me a little bit of the Mustafa Suleyman, which I mentioned probably every other episode, the September 2025 article on that. Mustafa Suleyman from Microsoft, head of Microsoft's AI, wrote this article called Seemingly Conscious AI, and his sort of warning or concern that if humans start to believe that AI is seemingly conscious, that could lead to a whole host of consequences, you know, some of which or many of which might not be particularly beneficial. But in all events, these bills are, and the one that you just mentioned, the Ohio HB469, are wading into this sort of cultural issue about people's romantic attachments to their AI chatbots. And it's really acknowledging when you're talking, and pulling out marriage as one of them, that people are concerned when humans are, which they are, trying to marry their chatbots, right? But the bill has a really interesting focus on accountability as well. It makes clear that AI systems can't own or control property, can't serve as an officer, director, or manager within a corporation, and that any harm caused by an AI system is the responsibility of the owner or user who directed the AI, not the AI itself. You know, in our tort system, in the our common law system of law and the way in which the tort doctrines have developed over now hundreds of years, both in the United States and England, we're always trying to figure out where does responsibility for harm lie. So, here the bill would basically make sure that the buck stops with a human. A human couldn't transfer property to an AI to try to shift and also cabin liabilities to the AI itself.
Scott Caravello: Yeah, and, I think, also, just to sort of connect it back to what you were discussing earlier, where, you know, a corporation's sole asset might be AI for those purposes, this kind of seems to be addressed squarely at that problem because, especially when you start talking about the fact that AI can't serve as the officer, director, or manager, you still would sort of… you would need a human to still do that, even if the sole asset was AI. So, we can't just really get at the AI corporation. But I think that that is a perfect segue into talking about the Argentinian president's proposal about AI personhood related to what he calls “automated companies” that are capable of operating autonomously through algorithms or AI systems without human employees. And Javier Milei, the president of Argentina, his basic argument is that the corporation, which limits liability for directors, officers, shareholders, and employees, enabled major economic innovation. And now for AI, we need something similar in the form of what he calls the non-human corporation.
Katherine Forrest: Right. And he published a piece in the Financial Times on this proposal arguing that for these new entities operated by AI agents or even robots to be genuinely useful, liability has to be limited. And so this is really a different outlook than what we're seeing in the US's state proposals that are focused on human accountability and keeping AI out of corporate positions. But, you know, the Financial Times in that article also published a rebuttal op-ed by Harari. You know, you've all Noah Harari, am I pronouncing his name right? You know I have a hard time with names, right, right? He's the Israeli historian philosopher. His books are fantastic. You know, Sapiens, a brief history of humankind. Everybody should read them. And Harari warned that granting legal status to corporations run by non-human agents could actually hand AI a master key to financial, economic, and political systems. And, in some ways, that's counting on additional intellectual capabilities of AI, one might call it superintelligence. And he also warned that countries recognizing AI-run corporations could become an AI state where inhabitants are governed by non-human corporations. So those dueling Financial Times articles, really worth taking a look at.
Scott Caravello: I totally agree. And I think that his critique is also so interesting, especially after all the episodes that we've done on agentic systems and cybersecurity, right? When agents act and coordinate, hold credentials, and have broad access, we know that things can go wrong. And so if we start to give agent corporations this legal protection and limit liability, what does the recourse look like? And going back to talking about human corporations, or rather corporations that are run by humans, when they get sued or prosecuted, we understand that there are still people who in some way might be held accountable, even if it comes from other mechanisms besides what's actually happening in the courtroom. Maybe someone gets replaced, maybe their compensation takes a hit if the stock price drops. But are we gonna be satisfied if in the agent corporation, the responsible AI agent just gets shut down and a new one gets spun up to take its place? It raises really incredible questions about what we're comfortable with and how we set up a functioning system for this sort of agentic corporation.
Katherine Forrest: Right. And I that this all leaves argument room for argument for doing a, sort of, “wait and see,” which is you don't have to attack the possibility of legal personhood now. You don't have to ban it forever. You don't have to grant it. That certainly would be premature. I think we can wait and see. And see where accountability is going to go, to see where the models are going to go, how different corporations are going to be setting up their own liability regimes. We're only starting to see some of the cases that are seeking to hold AI models and their developers accountable through their corporate entities now. And there's a ways to go before we actually figure out how that is best done.
Scott Caravello: Completely agree. But, so, what is the practical takeaway of all this and where we are right now? The near-term law is mostly about accountability. Idaho, Utah, Tennessee are just saying that AI can't be a legal person. And Ohio, and some others like Missouri, would have gone further by declaring AI non-sentient and then tying responsibility back to humans and developers. And then meanwhile, Argentina, and you know, this is really just from a week or so ago. So, it is new and we'll see where it goes and how the discussion evolves around it. But they're testing a different direction by imagining AI-operated companies as a strategy for economic growth.
Katherine Forrest: Right. And the and the deeper takeaway, you know, if there is one, and I think there is one, is that personhood is not a magic word. It's a legal tool. It's one that's been used before with corporations. We've had non-human persons for a very long time, so we shouldn't get just sort of, you know, scared and freaked out by the concept of personhood. We should think about it as a potential legal tool that can be used to establish roles and responsibilities in in a variety of different ways. And if AI remains merely a tool, you know, maybe we just end up regulating the humans and the companies behind it. But if AI becomes something more, for instance, super intelligent, as I have been suggesting, as you know, Mr. Caravello, then I think there's a reason to expect that you know, it might in some way, shape or form, actually require more from our legal system and we're gonna then have to be prepared to answer the harder questions. But I think we can take a wait and see attitude for the moment.
Scott Caravello: I agree, but you know, it is 10:58 AM. That is a lot to take in for a Tuesday morning recording.
Katherine Forrest: Well, you know, it's only because we're all still sleep deprived from Knicks and post-Knicks sort of festivities, right? Right?
Scott Caravello: Exactly
Katherine Forrest: Okay. All right. And we like to give a lot and a little bit of time here on Waking Up with AI. And that's all we've got time for. I'm Katherine Forrest.
Scott Caravello: And I'm Scott Caravello. Don't forget like and subscribe.
