The Frontier: Person, Product, or Something New?

A person for what?

Is a chatbot a person? Put that way, the question has no legal answer yet, and the law may never need one. Law rarely asks whether something is a person in the abstract. It asks narrower questions: whether a thing can own property, sign a contract, sue, be sued, or answer for harm. It answers each one for its own reasons, and it has given yes to some odd candidates and no to some sympathetic ones.

So this reading makes one claim and follows it through. In law, personhood is a tool the law picks up when it does a job. So far the law has handed AI systems its two other categories for things, product and tool, and whether it needs a new one is the part that is not settled. That open question is where the activity ends the course. You follow one live matter to its record and say what the law in it should call the machine.

Persons are made, not found

Florida’s statutes open with a list of definitions, and “person” is one of them. Section 1.01(3), Florida Statutes, says the word covers individuals and children, and also firms, partnerships, trusts, estates, corporations, and “all other groups or combinations.” Most of that list cannot think. A corporation is a set of papers filed in Tallahassee. Yet under section 607.0302 (opens in a new tab) of the same statutes, it has the same powers as an individual to carry on its business, including the power to sue and be sued and to own property. It can also last forever. The reasons are practical, and the video under the activity’s title walks through some of them. Investors wanted to pool their money for ventures that might fail, and to risk only what they put in. The law gave the corporation legal personhood because that solved the problem.

Admiralty law went further with a ship. In United States v. The Brig Malek Adhel, 43 U.S. (2 How.) 210 (1844) (opens in a new tab), the Supreme Court upheld the forfeiture of a brig whose captain had fired on other vessels at sea. Its owners had no part in the attacks, and the Court accepted that they were innocent. It made no difference. The vessel “is treated as the offender,” Justice Story wrote, whatever its owners did or knew. No one thought the ship could think. Treating it as the wrongdoer let the government reach the thing that did the harm when the people behind it were out of reach.

The law also says no. A crested macaque named Naruto was said to have taken the photographs known as the monkey selfies, and an animal-rights group sued in his name for copyright infringement. In Naruto v. Slater, 888 F.3d 418 (9th Cir. 2018) (opens in a new tab), the Ninth Circuit held that the monkey could not sue under the Copyright Act, because the Act does not authorize animals to bring those suits. Florida said no to a creek. In 2020 Orange County amended its charter to give the county’s waters a right to exist and to flow. That same year the Legislature barred any local government from granting legal rights to “a body of water” or any other part of the natural environment. A group of lakes and streams sued a developer and the state’s environmental agency under the charter. In Wilde Cypress Branch v. Hamilton (opens in a new tab), the Sixth District held the charter provision preempted, and the suit failed.

In other words, the law makes a person when making one serves a purpose, and it refuses when refusing serves one. None of these four calls turned on whether the thing could think or feel. A corporation cannot feel anything, and it can sue. A monkey can feel a great deal, and it cannot sue under the Copyright Act.

What the law calls AI today: a product or a tool

Asked about AI, courts and legislatures have so far reached for the categories the law keeps for things. The clearest ruling is Thaler v. Perlmutter (opens in a new tab). A computer scientist listed his AI system as the only author of a picture and asked for a copyright. The D.C. Circuit held that the Copyright Act requires a human author, and its reasons came from the statute’s own words. Authors own property, have lifespans and heirs, sign transfers, and form intentions. Wherever the Act mentions machines, the court said, it treats them as “tools, not authors.” The Supreme Court declined to hear the case in March 2026, so for copyright the point is settled. The court was careful about the rule’s limits: it does not bar a copyright in work a person makes with an AI system’s help.

The two categories that remain, product or tool, answer different questions. A product can be defective, and its maker answers for the defect. In 2025 a federal judge in Orlando, ruling on motions to dismiss in Garcia v. Character Technologies, Inc. (opens in a new tab), treated a chatbot app as a product for the claims about its design. The order drew a line, though: it did not treat the app as a product for the ideas and expressions inside it (page 36). A tool, by contrast, acts for whoever uses it. Florida’s version of the Uniform Electronic Transaction Act, section 668.50, Florida Statutes (opens in a new tab), calls a program that acts “without review or action by an individual” an electronic agent. Two such agents can form a contract that no person ever read, and the contract binds the parties whose agents they are.

Some legislatures want the answer written into statute. Idaho’s code has said since 2022 that artificial intelligence “shall not be granted personhood” (opens in a new tab) in the state. Ohio’s House Bill 469 (opens in a new tab) would go further. It would declare AI systems nonsentient, void any attempt to marry one or seat one on a corporate board, and bar one from owning property. For harm an AI system causes, it would hold responsible the owner or user who directed it, and the maker for defects under the principles of product liability. Read closely, the bill is the tool and product answers put into one chapter of a code.

The case for something new

The old categories strain at the edges, and the argument lives there. Those who want a third category point to two strains. A product rule looks for a defect someone built in, but a system that writes new words for every user can cause a harm no one built in. A tool rule treats the machine’s act as the user’s. But when a program acts for hours without anyone reviewing its steps, “the user did it” is harder to believe than it is for a hammer. Those who resist a third category answer that both rules still put a person on the hook, and that a person is who belongs there.

In 2017 the European Parliament asked the European Commission to study electronic personhood, a status under which the most sophisticated autonomous robots could answer for damage they cause. The resolution (opens in a new tab) listed it beside compulsory insurance for robots and a compensation fund, two options that keep the cost on people. The idea went no further. The AI Act the EU adopted in 2024 puts its duties on the people and companies that build and use AI, never on the systems.

The loudest talk of something new has come from outside the law. One developer trains its models on a written constitution (opens in a new tab) that calls such systems “a genuinely new kind of entity” and treats their moral status as uncertain. Idaho’s lawmakers have written the opposite view into law, and Ohio’s bill would do the same. A document like that constitution still governs a model before any statute does. Lawrence Lessig’s four modalities of regulation explain how: law is one constraint among four, beside social norms, the market, and architecture. His chapter on what things regulate (opens in a new tab) shows the four acting on a single person’s choice. A company that builds its values into a model through training regulates by architecture, whether or not a legislature ever speaks.

Meanwhile the law moves at its own speed. A court decides the case in front of it, with the law that exists. A legislature meets on a calendar, and an agency takes comments before it acts. That lag is the pacing problem, and it means a new kind of thing gets the old categories until someone writes a new one. Product and tool are the law’s answers for now. Whether they are its last word is not settled.

Following what is not settled

Every matter on the activity’s watch list sits somewhere in this argument. A court deciding whether ChatGPT practiced law is asking whether a program can do what only a licensed person may. A judge weighing an AI shopping agent’s visits to a store’s website is asking whose act those visits were. A bill on digital replicas is asking who controls a voice a machine can copy. None of those questions will be answered in a textbook first.

To follow one, start where it happens. Reading the record means opening the docket, the bill’s history, or the committee’s file and finding the last dated event. A news story can lead you there, but it reports where the matter stood when it was written, which is the difference between a development and a headline. When the story and the record disagree, the record wins, and your brief says what the record says.

Then say it plainly. Explaining to a general audience takes one point at a time, no fact you have not checked, and an ending that tells the listener what to watch. Your brief closes by asking which category the law now uses for the machine in your matter, and whether it fits. That answer is yours to argue. This reading claims only that the question is open, and that the record is where it will be decided.

Start with Part 1 of the activity: choose your item, and before you write a word, find out whether it has moved.