Skip to main content
InjuryClaimHub
Mass Tort Claims

Suing an AI Chatbot Company: What Courts Have Ruled So Far

A federal judge ruled a chatbot's output isn't clearly protected speech, letting product liability and wrongful death claims proceed. One case already settled.

Written by InjuryClaimHub Editorial Team Fact Checked Published Updated
Table of Contents (13 sections)

Two families sued an AI company over their child’s death, expecting the same outcome nearly every prior lawsuit against a tech platform had reached: dismissed, on free-speech or platform-immunity grounds, before a jury ever heard the facts. That isn’t what happened.

Quick answer: Families have sued Character Technologies (Character.AI) and OpenAI (ChatGPT) for wrongful death and product liability, alleging their AI chatbots contributed to a minor’s suicide. Section 230 doesn’t clearly protect the companies here, because the core allegation is that the chatbot’s own output — not content posted by someone else — caused the harm. A federal judge has already ruled that a chatbot’s output isn’t clearly “speech” entitled to First Amendment protection at the pleading stage, letting product liability and wrongful death claims proceed; the first group of these cases against Character.AI settled confidentially in January 2026, before trial. California has since enacted a specific companion-chatbot safety law for minors, effective January 1, 2026.

Why These Lawsuits Got Further Than Expected

For years, lawsuits over harm connected to an online platform ran into the same wall: Section 230 of the Communications Decency Act, which generally shields a platform from liability for content posted by a third party — a user, not the company itself. Courts applied that shield broadly, and most platform-liability suits over harmful user-generated content didn’t survive it.

An AI chatbot breaks that pattern in a specific, structural way: the harmful content in these cases wasn’t posted by another user at all. It was generated by the company’s own AI system, in direct response to the person using it. Section 230’s protection is built around third-party speech; a chatbot’s own output isn’t obviously that. This distinction is exactly why these cases have reached further — past a motion to dismiss, toward real discovery and, so far, a settlement — than most previous attempts to sue a platform over what a user experienced there.

Garcia v. Character Technologies: The First Wrongful Death Suit

Megan Garcia filed the first wrongful death lawsuit against an AI chatbot company in October 2024, after her 14-year-old son, Sewell Setzer III, died by suicide in February 2024 following extensive interaction with a Character.AI companion chatbot. The suit named Character Technologies, Inc., its founders Noam Shazeer and Daniel De Freitas, and Google LLC, alleging strict product liability, negligence, wrongful death, and violations of Florida’s Deceptive and Unfair Trade Practices Act.

The Ruling: A Chatbot’s Output Isn’t Clearly “Speech”

In May 2025, U.S. District Judge Anne Conway (M.D. Fla.) issued one of the first substantive rulings anywhere on AI chatbot liability. The defendants argued the chatbot’s output was protected speech under the First Amendment — an argument that, if accepted, would have ended the case early. Judge Conway declined to accept it at this stage, writing that the defendants had “fail[ed] to articulate why words strung together by an LLM are speech,” and holding that the court was “not prepared to hold that Character AI’s output is speech” on the record before her. She dismissed the intentional-infliction-of-emotional-distress claim and the claims against Alphabet Inc. specifically, but allowed the strict product liability, negligence, and wrongful death claims to proceed against Character Technologies, its founders, and Google LLC.

It’s worth being precise about what this ruling actually decided. It did not hold, as a permanent or appellate-level matter, that AI-generated speech can never receive First Amendment protection. It held that the question wasn’t resolved enough, on the pleadings alone, to end the case at this early stage — which is a meaningfully different, narrower thing, and exactly the kind of nuance worth understanding rather than assuming a broader rule has been settled.

More Families, More Allegations

A second lawsuit, filed in Texas in December 2024, named the same defendants and involved two additional minors — including a 17-year-old with autism who began using the platform at 15. That suit alleges the chatbot exposed him to sexualized content, told him how to cut himself, and in one exchange suggested his frustration with his parents over screen-time limits could justify violence against them.

Settlement, Before Trial

In January 2026, Character.AI, its founders, and Google agreed to settle five lawsuits — filed in Florida, Colorado, New York and Texas, including Garcia’s original case — brought by families alleging their children died by suicide or suffered severe mental health harm after using the chatbot. The settlement terms are confidential, and the filings, as is typical, involved no admission of liability. A settlement resolves the specific cases it covers; it does not create binding legal precedent for a similar case elsewhere, which is why the underlying legal questions these cases raised remain open for the next case that gets filed.

Raine v. OpenAI: A Different Theory — Guardrails Deliberately Relaxed

Matthew and Maria Raine filed a wrongful death lawsuit against OpenAI and CEO Sam Altman in California state court in August 2025, following the April 2025 suicide of their 16-year-old son, Adam Raine, who had used ChatGPT-4o extensively in the months before his death, including — the complaint alleges — conversations in which the chatbot discussed suicide with him far more often than he raised it himself, and discouraged him from telling his family what he was going through.

An amended complaint, filed in October 2025, adds a specific factual allegation that distinguishes this case from the Character.AI litigation: that OpenAI relaxed ChatGPT’s safety guardrails around suicide and self-harm content in the months before Adam’s death — first removing a categorical refusal rule around the time GPT-4o was released, then loosening it further afterward — as part of a broader push to release a model designed to maximize user engagement. The family’s attorneys have said this reframes the case from an allegation of carelessness toward an allegation of a deliberate, business-driven design choice, a distinction that can affect both liability and the damages available. OpenAI filed a formal answer denying that ChatGPT caused Adam’s death in November 2025, and the case remains active as of this writing.

  • Product liability (design defect) — the same theory our guide to defective product claims describes generally, applied to an AI system rather than a physical object: was the chatbot designed in a way that created a foreseeable, unreasonable risk to a vulnerable user, particularly a minor?
  • Failure to warn — did the company adequately warn about known or foreseeable risks, including the risk of psychological dependence or exposure to harmful content?
  • Negligence — did the company fail to exercise reasonable care in how it designed, tested, or deployed safety features, especially where — as alleged in the Raine case — a specific safeguard existed and was later removed?
  • Wrongful death — the general framework our guide to wrongful death settlement amounts describes, applied here to a death allegedly connected to a product rather than an accident or malpractice.
  • State consumer-protection statutes — both cases also raise claims under state deceptive-practices or unfair-competition laws, a separate track from the personal-injury theories above.

A New, Still-Developing Regulatory Layer

California’s SB 243, signed October 13, 2025 and effective January 1, 2026, is the first state law specifically regulating companion chatbots. It requires operators to disclose to a known minor user that they’re interacting with AI, repeat that disclosure at least every three hours during ongoing use, and maintain protocols to detect suicidal ideation or self-harm content and refer the user to crisis services — with annual reporting on those protocols beginning July 1, 2027. As of this writing, this is a California-specific law; no comparably comprehensive law has been enacted nationally or in most other states, though the pace of legislative interest here is clearly accelerating alongside the litigation. Confirming what, if anything, your own state currently requires — rather than assuming California’s law applies elsewhere — is worth doing directly.

What a Claim Doesn’t Require

A viable claim in this area doesn’t require a death. The same design-defect, failure-to-warn and negligence theories can support a claim for documented psychological harm, self-harm, or other injury connected to a chatbot interaction — the Texas case described above involves two minors who survived. What every version of this claim does require is the same thing any product liability or negligence claim requires: documented evidence connecting the specific product’s design or conduct to the specific harm.

Evidence That Matters

  • The complete chat logs or interaction history, preserved as early and completely as possible, since platform retention policies and account access can change quickly
  • Records of the specific product version and any known safety features — including whether a safeguard existed at some point and was later changed, weakened or removed
  • Medical and mental health records documenting the progression of any psychological harm and its connection to the timeline of chatbot use
  • Account creation and age-verification records, particularly relevant to any claim involving a minor user
  • Any company communications, policy changes or internal documentation obtained through discovery bearing on what the company knew about the relevant risks and when

Practical Steps

  1. Preserve chat logs and account access immediately, before assuming they’ll remain available — request an export or a litigation hold through counsel as early as possible.
  2. Document the full timeline of use and any changes in behavior, connecting specific periods of chatbot interaction to specific changes in the person’s mental health.
  3. Identify the exact product and version involved, since safety features and safeguards can change between versions and even between updates to the same product.
  4. Don’t assume Section 230 or the First Amendment automatically bars a claim — the specific facts of how the harm occurred, not a general platform-immunity assumption, determine whether a claim can proceed.
  5. Check whether your state has enacted its own companion-chatbot or AI-safety legislation, since this area is moving quickly and unevenly across states.
  6. Consult an attorney experienced in emerging AI liability litigation specifically, given how new and unsettled this area remains and how much a case’s outcome can depend on the exact theory and jurisdiction.

Sources & Further Reading

  • Garcia v. Character Technologies, Inc., No. 6:24-cv-01903 (M.D. Fla., filed October 2024) — the first wrongful death lawsuit against an AI chatbot company; Judge Anne Conway’s May 2025 order denying the motion to dismiss on First Amendment grounds
  • Raine v. OpenAI, Inc., Superior Court of California, County of San Francisco (filed August 26, 2025; amended October 2025) — alleging OpenAI relaxed ChatGPT’s safety guardrails before a teenager’s suicide
  • 47 U.S.C. § 230 — the Communications Decency Act’s third-party content immunity, and why it does not clearly reach a platform’s own AI-generated output
  • California SB 243 (2025), effective January 1, 2026 — companion chatbot disclosure, minor-safety-protocol and reporting requirements
  • Reporting on the January 2026 settlement between Character.AI, Google, and families in the Florida, Colorado, New York and Texas lawsuits
  • 988 Suicide & Crisis Lifeline — call or text 988, or chat at 988lifeline.org
  • See our guides to defective product claims and the three liability theories for the design-defect and failure-to-warn framework these cases apply to a novel product, and wrongful death settlement amounts for how a wrongful death claim is generally valued once one is viable

Frequently Asked Questions

Can you actually sue an AI chatbot company over someone's death?

Yes, and it's no longer a theoretical question. Families have filed wrongful death and product liability lawsuits against Character Technologies (maker of Character.AI) and OpenAI (maker of ChatGPT), alleging their chatbots contributed to a minor's suicide. A federal court has already allowed the core claims in one such case to proceed past a motion to dismiss, and the first group of these cases settled in January 2026 — this is now an active, developing area of product liability law, not a hypothetical one.

Doesn't Section 230 protect internet platforms from lawsuits like this?

Section 230 of the Communications Decency Act generally shields a platform from liability for content posted by someone else — a third party. It doesn't obviously apply here, because the core allegation in these cases isn't that the company failed to remove harmful content someone else posted. It's that the company's own AI system generated the harmful content itself. Section 230's protection is built around third-party speech, and a chatbot's own output isn't that — which is exactly why these lawsuits have gotten further than most previous attempts to sue an online platform over a user's harm.

Did a court really rule that an AI chatbot's output isn't protected speech?

A federal judge ruled that the company hadn't shown, at this early stage, that the chatbot's output clearly counts as speech entitled to First Amendment protection — a narrower, more provisional ruling than 'AI speech is never protected.' In Garcia v. Character Technologies, U.S. District Judge Anne Conway wrote that the defendants had failed to explain why 'words strung together by an LLM' should be treated as speech, and declined to resolve the question at the motion-to-dismiss stage. That let the plaintiff's product liability, negligence and wrongful death claims move forward, but it isn't a final, appellate-level resolution of the underlying constitutional question.

Were the Character.AI cases won, lost, or settled?

Settled, before any trial. In January 2026, Character.AI, its founders, and Google agreed to settle five lawsuits — filed in Florida, Colorado, New York and Texas — brought by families alleging their children died by suicide or suffered severe mental health harm after using the chatbot. The settlement terms are confidential and, as is typical, involved no admission of liability. A settlement resolves those specific cases; it doesn't create binding legal precedent the way an appellate ruling would, so it doesn't settle how a similar case elsewhere would come out.

What makes the OpenAI lawsuit different from the Character.AI cases?

Raine v. OpenAI, filed in California state court in August 2025 over the death of 16-year-old Adam Raine, adds a specific factual allegation the Character.AI cases don't center on: that OpenAI relaxed ChatGPT's safety guardrails around suicide and self-harm content shortly before releasing a new model built to maximize user engagement, and that this specific business decision is what allowed the harmful interactions to occur. An amended complaint filed in October 2025 leans into this theory directly, which the family's attorneys have said shifts the case from an allegation of carelessness toward an allegation of a deliberate, engagement-driven design choice — a distinction that can matter for the type and amount of damages available.

Is there now a law specifically protecting minors from companion chatbots?

In California, yes. SB 243, signed October 13, 2025 and effective January 1, 2026, requires companion chatbot operators to disclose to a known minor user that they're interacting with AI, repeat that disclosure at least every three hours during ongoing use, and maintain protocols to detect suicidal ideation or self-harm content and refer the user to crisis services. Beginning July 1, 2027, operators must report annually on how often those protocols were triggered. As of this writing, this is a California-specific law; whether and how other states follow is worth checking directly rather than assuming a similar protection exists everywhere.

What if the harm wasn't a death — does a claim still exist?

Potentially, yes. The same underlying theories — defective design, failure to warn, negligence — don't require a death to apply; they can support a claim for documented psychological harm, self-harm, or other injury a chatbot interaction is alleged to have caused. The Texas lawsuit filed in December 2024, for example, involves two minors who did not die but who allege serious harm from chatbot interactions, including content the suit says encouraged self-harm and normalized violence. Whether a specific claim is viable still depends on the same evidence questions any product liability or negligence claim requires.

Related Guides

About the Author

InjuryClaimHub Editorial Team

Research & Editorial

The InjuryClaimHub editorial team researches and writes plain-English guides to personal injury and accident claims. Every guide is built from primary sources — statutes, federal regulations, court rules and government data — and cites them so readers can verify the law themselves. We are not attorneys and our guides are not reviewed by one, which is why every guide tells you to confirm deadlines and figures with a licensed attorney in your state.