Does Your AI Have a Right to Free Speech?

When an AI chatbot says something harmful, like defaming a real person or revealing private information about someone who never agreed to be exposed, who answers for it? More and more, AI companies have a ready answer: nobody. What the chatbot says, they argue, is protected by the First Amendment.

That’s a big claim. And it raises a question courts are only starting to work through: if an algorithm, not a person, decides what to say, is that speech the Constitution protects?

Who Is Doing the Talking?

The First Amendment protects people. Corporations can claim free speech rights too, but only because they are made up of people. As Justice Scalia put it, “the power to publish thoughts, no less than the power to speak thoughts, belongs only to human beings.” (Citizens United v. FEC, 558 U.S. 310 (2010) (Scalia, J., concurring).)

"Machines today remain like Blackie the cat, property of their masters and owners. They remain, to the law at least, dumb beasts."

A chatbot is not a person. When the companies behind Character.AI, including Google, argued that their chatbot’s output was protected speech in a lawsuit over a teenager’s death, a federal court in Florida declined to go along, noting that a chatbot “is not a ‘person’ and is therefore not protected by the Bill of Rights.” (Garcia v. Character Technologies, Inc., 785 F. Supp. 3d 1157 (M.D. Fla. 2025).) To support that point, the court cited an earlier case rejecting a free speech claim brought on behalf of a talking cat. Columbia law professor Tim Wu made the same comparison over a decade ago: “machines today remain like Blackie the cat, property of their masters and owners. They remain, to the law at least, dumb beasts.” (Tim Wu, Machine Speech, 161 U. Pa. L. Rev. 1495 (2013).)

So AI companies make a different argument. The speech isn’t the chatbot’s, they say. It’s theirs.

The Algorithm Problem

That argument leans heavily on Moody v. NetChoice, where the Supreme Court held in 2024 that Facebook and YouTube’s feeds were protected because they reflected content moderation choices made by people at those companies. (603 U.S. 707 (2024).) But the Court was careful to limit that holding. Justice Barrett wrote separately to ask what happens when a company “hand[s] the reins to an AI tool,” and warned that technology “may attenuate the connection” between what appears on screen and any human’s protected choices. Justice Alito went further, observing that “even the researchers and programmers creating” AI models “don’t really understand why the models they have built make the decisions they make.”

The Ninth Circuit has since picked up on that point, explaining that whether an algorithm is expressive depends on how it actually works, and that algorithms that simply serve users what they’ve shown they want are “probably not expressive.” (NetChoice, LLC v. Bonta, 152 F.4th 1002 (9th Cir. 2025).)

That’s the heart of the problem. A newspaper editor chooses what to print. A chatbot’s makers don’t choose what it says in response to any given question. Often, they couldn’t predict it if they tried.

What the Scholars Say

Legal scholars are divided. On one side, Eugene Volokh, Mark Lemley, Peter Henderson, and others argue that AI output should generally be protected, mostly because of the rights of the people who use it, including their right to receive information. (Freedom of Speech and AI Output, 3 J. Free Speech L. 651 (2023).) Toni Massaro and Helen Norton similarly observe that current free speech doctrine poses “surprisingly few barriers” to protecting machine speech. (Siri-ously? Free Speech Rights and Artificial Intelligence, 110 Nw. U. L. Rev. 1169 (2016).)

“AI, as such, does not have First Amendment rights, just as televisions, hats, motor vehicles, and electric blankets do not have First Amendment rights.”

On the other side, Professor Peter Salib argues that “AI outputs are not protected speech,” because when a chatbot answers a question, no one with First Amendment rights is actually communicating. A chatbot’s output is no more its creator’s speech than a child’s words are her parents’. (AI Outputs Are Not Protected Speech, 102 Wash. U. L. Rev. 83 (2024).) A recent Stanford Law Review article makes a similar point: speech requires that the speaker know what they said when they said it, and machine-learning systems fail that test because even their programmers can’t identify what the system will say. (Mackenzie Austin & Max Levy, Speech Certainty: Algorithmic Speech and the Limits of the First Amendment, 77 Stan. L. Rev. 1 (2025).) And Harvard’s Cass Sunstein puts it bluntly: “AI, as such, does not have First Amendment rights, just as televisions, hats, motor vehicles, and electric blankets do not have First Amendment rights.” (Artificial Intelligence and the First Amendment, 92 Geo. Wash. L. Rev. 1207 (2024).)

And as Sunstein points out, whatever the answer, one thing is clear: speech that isn’t protected when a person says it doesn’t become protected because a machine said it instead. Defamation is still defamation.

Having It Both Ways

There’s one more wrinkle. When AI companies are sued over harmful content, they often argue two things at once. Under Section 230, they say the content wasn’t really theirs, so they can’t be held responsible for it. Under the First Amendment, they say the content is their own protected speech. It can’t be both.

Courts will be sorting this out for years. It's happening now in one of our cases (see the Case Brief here). But the First Amendment was written to protect human expression, not to give technology companies a free pass for whatever their machines produce.

If you’ve been harmed by something an AI chatbot said about you, talk to us. It’s our job to help you.