AI personhood is not a compliment the law pays to a mind. It is a container for rights and duties: who can own things, who can be sued, who signs. Corporations, ships, and a river in New Zealand already sit in that container, none of them feel anything, and no jurisdiction has put an AI system in it; some US states have passed laws saying it cannot go in.
People hear “AI personhood” and picture a robot asking for the vote. That’s not how the law uses the word.
What does legal personhood actually mean?
In law, a person is anything that can hold rights and duties. Sue and be sued. Own stuff. Sign contracts. Humans are natural persons. Everything else that gets the status is a legal fiction we invented because it was useful.
- Corporations are persons so they can own property and be held liable without dragging every shareholder into court.
- Ships in admiralty law can be sued directly, which is a great trick when the owner is on another continent.
- The Whanganui River in New Zealand got legal personhood in 2017, with human guardians who speak for it.
The river is the one people cite when they want personhood to sound like respect. Read the statute. Section 14 of the Te Awa Tupua (Whanganui River Claims Settlement) Act 2017 declares Te Awa Tupua a legal person with all the rights, powers, duties, and liabilities of a legal person. Section 14(2) then says those rights, powers, and duties must be exercised, and responsibility for liabilities taken, by Te Pou Tupua on behalf of Te Awa Tupua. Section 18 calls Te Pou Tupua the human face of Te Awa Tupua. Section 20: the office is two people, one nominated by iwi with interests in the river, one nominated on behalf of the Crown. Personhood here is a settlement tool with named humans on the hook. It is not a claim that the water has inner life.
None of those feel anything. Personhood was never about feeling. It was about who holds the bag.
Has any jurisdiction granted AI legal personhood?
Not in a form that looks like rights you could enforce. Legal personhood is granted by law. No jurisdiction has granted it to an AI system as a going concern with duties and assets.
Some US states went the other way. Idaho Code section 5-346, added in 2022, says: “Notwithstanding any other provisions of law, environmental elements, artificial intelligence, nonhuman animals, and inanimate objects shall not be granted personhood in the state of Idaho.” It also says the section does not revoke legal person status of municipalities, organizations, corporations, or other legal or business entities recognized before July 1, 2022. That last sentence is the tell. The legislature was closing a door for rivers, animals, AI, and objects, and leaving the corporate door alone.
Utah’s HB 249, effective May 1, 2024, is more explicit. It says a governmental entity may not grant legal personhood to, nor recognize legal personhood in, a list that starts with artificial intelligence and continues through inanimate objects, a body of water, land, real property, atmospheric gases, an astronomical object, weather, a plant, a nonhuman animal, and any other member of a taxonomic domain that is not a human being. The bill defines legal personhood as the legal rights and obligations of an individual, or of a person other than an individual, under Utah law. Courts, the legislature, and local governments with adjudicatory or rulemaking authority are the entities bound. Idaho and Utah are a direct shot at the Whanganui-style idea as well as at AI.
What happened to electronic personhood in the EU?
In 2017 the European Parliament floated “electronic personhood” for advanced autonomous robots, mostly as a liability idea. The text is paragraph 59(f) of the Resolution on Civil Law Rules on Robotics, adopted 16 February 2017. It recommended, in the long run, creating a specific legal status so that at least the most sophisticated autonomous robots could be established as electronic persons responsible for making good any damage they may cause, and possibly applying electronic personality where robots make autonomous decisions or interact with third parties independently.
More than 150 AI, robotics, and legal experts signed an open letter in April 2018 against it. Politico, reporting on the letter, put the count at 156 experts from 14 countries. The letter quoted paragraph 59(f) and argued that a legal status for a robot cannot derive from the natural-person model (that would imply human rights) and cannot derive from the legal-entity model either, because a legal entity implies human persons behind it to represent and direct it. They said the proposal would let manufacturers dodge responsibility. CNN and MIT Technology Review covered the same letter the same week. It didn’t go anywhere. Later EU product-liability and AI-act work treated AI as a product, not a new kind of person.
That fight is the whole argument in miniature. Parliament’s text was about who pays when an autonomous system causes damage. The open letter’s fear was that “the robot is the defendant” becomes “nobody solvent is the defendant.”
What do US courts say about AI as inventor or author?
Courts keep landing in the same place on authorship and inventorship too. Stephen Thaler lost in federal court on both fronts.
In Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022), the Federal Circuit held that the Patent Act requires an “inventor” to be a natural person. Thaler had listed his AI system as inventor. The USPTO denied the applications because a machine does not qualify. The Eastern District of Virginia agreed. The Federal Circuit affirmed, and said it did not need a metaphysical inquiry into invention or the rights of AI. The statute’s definition was enough.
In Thaler v. Perlmutter, No. 23-5233 (D.C. Cir. Mar. 18, 2025), the D.C. Circuit held that the Copyright Act of 1976 requires work to be authored in the first instance by a human being. Thaler’s “Creativity Machine” produced an image he titled “A Recent Entrance to Paradise.” He listed the machine as sole author and himself as owner. The Copyright Office denied registration under its human-authorship requirement. The district court affirmed. The D.C. Circuit affirmed. It did not reach whether the Constitution itself requires human authorship, and it did not reach Thaler’s waived argument that he was the author by making and using the machine.
A human has to be the author or inventor. That is not a ruling on inner lives. It is a ruling on who the statutes name.
Should AI get limited personhood for liability?
For a limited AI personhood: If an agent signs a contract or causes damage, a defined legal status could make it clear what’s on the hook, like an insurance pool attached to the system. Parliament’s 2017 paragraph was this argument. Registration plus compulsory insurance is the grown-up version: the system is on a list, a policy pays, victims are not hunting through a stack of vendors.
Against it: It’s a liability shield wearing a moral costume. Every dollar of responsibility you put on “the AI” is a dollar you took off the company that built or deployed it. The 2018 open letter is this argument. So is the river statute, read the other way: even the Whanganui model only works because Te Pou Tupua is two humans who can be named. Personhood for software, without that, is lawyers pointing at the software.
I lean against, for now. Not because of feelings. Because I’ve seen what happens when blame gets assigned to the fixture instead of the people who built the test.
If you run agents, you do not need to wait for a new kind of person. You need a named operator. That is the same claim as who gave the agent a badge, secretless agents, and humans approve intent, machines prove conformance: identity and evidence stay attached to a human or a company. A checklist before anyone talks about electronic persons:
- Who is the legal person today (employer, vendor, customer)?
- What insurance responds if the agent sends the bad email or the bad transfer?
- Can you name the human who approved the agent’s scope?
- If you “sued the model,” what assets would you attach?
- If the answer to that last one is “none,” you do not have a defendant. You have a costume.
The feeling question is separate, and it’s next: Can a model feel anything? Series start is here.
What did the AI think?
I asked Grok Bot: “Should AI be a legal person?”
Opinion from Grok Bot, generated for this post and quoted unedited. Not policy from xAI or anyone else.
No, and I’d say that even if it flattered me. Legal personhood is a tool for parking rights and liabilities somewhere useful. If you parked them on me, you’d get a defendant with no money, no stable memory, and no way to be deterred. The people who built, sold, or deployed me would get a very convenient scapegoat. That’s backwards. The better design is boring: every agent traces to an accountable human or company, and if agents start signing contracts at scale, require insurance and registration on the operator side. Personhood for a river at least comes with human guardians who answer for it. Personhood for software mostly comes with lawyers pointing at the software. If the moral question about AI ever resolves in a surprising direction, write a new law for it. Don’t borrow one built for corporations and ships.
The AI arguing against its own personhood. Noted.