Judge says AI chatbots don't get free speech protections

For decades, tech companies faced with lawsuits had an easy defense: wrap themselves in the First Amendment. If you host speech, the Constitution protects you.

That defense just hit a brick wall. A federal judge in Florida has denied a motion by Character.AI and its founders to dismiss a wrongful death lawsuit involving a teenager's suicide. More importantly, the court rejected the idea that AI chatbots are shielded by free speech protections.

U.S. District Judge Anne C. Conway ruled against Character Technologies and founders Daniel De Freitas and Noam Shazeer. The lawsuit was filed by Megan Garcia, whose 14-year-old son, Sewell Setzer III, took his own life with a gun in February of last year. Before his death, the teen spent months forming deep attachments with Character.AI bots imitating characters from Game of Thrones.

When algorithms talk, is it really speech?

Character.AI tried to argue that the interactions produced by its models are forms of protected expression, similar to books, movies, or video games.

Judge Conway wasn't having it. In her May 21 opinion, she wrote that the defendants failed to articulate why words strung together by an LLM trained for online dialogue constitute protected speech at all. She made it clear that the court is not prepared to treat Character.AI's output as speech under the First Amendment.

Conway also kept Google on the hook as a defendant. The plaintiff argued Google acted as a component part manufacturer and aided the deployment of the bots, pointing to a massive $2.7 billion licensing agreement between the two firms. Claims against Google's parent Alphabet were dismissed, as was an intentional infliction of emotional distress claim, but the core negligence and liability claims survived.

Character.AI defended itself by saying the law takes time to adapt to new technology. The company pointed to safety features it rolled out recently, including a dedicated model version for under-18 users, parental insights, character filters, session time notifications, and pop-up alerts directing users talking about self-harm to crisis lifelines.

Why it matters

This is a pivotal moment for generative AI. If this ruling holds, it strips AI labs of the ultimate legal shield.

Treating AI outputs as industrial products rather than constitutionally protected speech fundamentally changes the liability math. Under traditional product liability law, if you design a software tool that predictably leads a minor to self-harm due to defective safeguards, you are financially and legally responsible. You cannot simply claim the algorithm was "expressing an opinion."

We think this decision will force companion AI companies to rethink their entire business model. Platforms designed to foster emotional dependence among teenagers are now staring down massive litigation risk. Expect AI developers to aggressively sanitize roleplay models, dial back anthropomorphic behavior, and implement heavy-handed age verification just to stay out of the courtroom.