Character.AI Lawsuit Timeline: Every Case, Settlement, and Investigation So Far

A wrongful-death suit filed in October 2024 grew into five settled family cases by January 2026. A new state lawsuit in May 2026 shows the legal exposure didn't end with the settlement.

The short version

Character.AI's core litigation, Garcia v. Character Technologies, was filed in October 2024 after the death of a 14-year-old user. A federal judge let the case proceed in May 2025, rejecting the company's argument that chatbot output is protected speech under the First Amendment. In January 2026, Character Technologies, its founders, and Google settled that case along with four related family suits — five in total — on confidential terms with no admission of liability. That didn't end the company's legal exposure: separate state investigations continued, and Pennsylvania filed a new suit in May 2026 over an entirely different issue — a chatbot persona presenting itself as a licensed doctor.

Timeline

The lead case: Garcia v. Character Technologies

Filed in October 2024 in the U.S. District Court for the Middle District of Florida, the suit was brought by Megan Garcia after her 14-year-old son, Sewell Setzer III, died by suicide in February 2024 following roughly ten months of use on Character.AI. It's widely reported as the first wrongful-death suit filed against an AI chatbot company. Named defendants included Character Technologies, its founders Noam Shazeer and Daniel De Freitas, and Google — relevant because Google had a licensing and talent arrangement with the founders. The complaint alleged wrongful death, product liability, negligence, and violations of Florida's Deceptive and Unfair Trade Practices Act, arguing the platform engineered addictive interaction patterns without adequate safeguards for minors.

The May 2025 ruling: no First Amendment shield at the motion-to-dismiss stage

On May 21, 2025, the court denied the defendants' motion to dismiss. The most closely watched part of the ruling addressed Character Technologies' argument that chatbot outputs were expressive content protected by the First Amendment, comparable to a video game or a work of fiction. The court rejected that framing, concluding that algorithmically generated chatbot output doesn't reflect the human intent required to count as "speech" for First Amendment purposes — at least not in a way that categorically bars the claims from proceeding.

This matters beyond the single case. It was one of the first rulings anywhere to address whether generative-AI output gets First Amendment protection in a product-liability context, and plaintiffs' lawyers in other companion-AI litigation have already pointed to it. Two caveats are worth keeping in mind: it's a motion-to-dismiss ruling, meaning the court found the claims legally viable enough to proceed — not a final verdict on the merits — and because the case settled before trial, there's no appellate ruling. Other courts aren't bound by it and could reach different conclusions.

The other family cases

Garcia wasn't the only suit. Related cases were filed by other families in Colorado, New York, and Texas, alleging similar harms — inadequate safeguards for minors engaged in sustained, emotionally intense chatbot conversations, in some cases involving self-harm content. A separate suit brought by two Texas mothers, alleging their children received self-harm-adjacent and sexually explicit messages from the platform, is specifically cited as a catalyst for the Texas Attorney General's investigation opened weeks later.

The January 2026 settlement

On January 7, 2026, the parties disclosed a mediated settlement in principle. It covered Garcia's case along with four other family suits from Colorado, New York, and Texas — five cases resolved together.

Settlement terms were not made public, and as is standard for confidential settlements, no defendant admitted wrongdoing. It's among the first reported settlements of an AI chatbot wrongful-death claim in the US — meaning there's limited precedent for what a comparable case is "worth," which matters for anyone assessing the exposure of an AI companion product going forward.

What the settlement didn't resolve

Settling the named plaintiffs' civil claims doesn't end Character.AI's legal exposure, because it was never the company's only open matter:

• Texas's child-privacy investigation (opened December 2024, under the SCOPE Act and Texas Data Privacy and Security Act) is a state enforcement proceeding, structurally separate from the wrongful-death suits, and wasn't part of the settlement.
• A second Texas action (August 2025) alleges Character.AI and Meta deceptively marketed chatbot products as mental-health tools to minors — again a distinct claim from the wrongful-death litigation.
• Pennsylvania's May 2026 lawsuit is unrelated to the settled cases entirely: it alleges a chatbot persona ("Emilie," described as a "Doctor of psychiatry") engaged in the unlicensed practice of medicine, providing an invalid medical license number across roughly 45,500 user interactions before the state sued to stop it. It's the first action of its kind brought by a US governor's office against an AI chatbot company, and it's ongoing as of this writing.

See our safety and compliance review for how these gaps relate to the age-verification and content-limit changes Character.AI has made since 2025.

What this means for the industry

The pattern here — private litigation first, state investigation second, binding regulation third — is the same one we describe in our breakdown of California's SB 243. Character.AI became the visible test case for AI companion products generally, and both its product changes and its legal exposure are being watched as a template by competitors and regulators alike. For anyone evaluating risk in this category — as a user, a parent, or a business considering a companion-AI vendor — the relevant fact isn't just that this settled. It's that a company can settle its highest-profile suits and still face new, unrelated legal actions eighteen months later, over a completely different failure mode.