What do the Character.AI and companion chatbot lawsuits mean for an AI app’s insurance?

What has been filed, and what a court has actually held

In Garcia v. Character Technologies, filed in federal court in Orlando in October 2024, a mother sued the maker of the Character A.I. app, its founders and Google over her 14-year-old son’s death. Everything about the underlying events is an allegation. What exists as a ruling is the court’s order of May 21, 2025 on the defendants’ motions to dismiss, which took the complaint’s facts as true.

On that footing the court held it was not prepared to treat the app’s output as speech at that stage, and it held the app is a product for the plaintiff’s product liability claims so far as they arise from defects in the app rather than from ideas or expressions in it. It let negligence, negligence per se, failure to warn, deceptive practices and unjust enrichment go forward, and dismissed the emotional distress claim. It also found a component part manufacturer theory against Google sufficiently alleged.

In Raine v. OpenAI, filed in San Francisco Superior Court on August 26, 2025, the parents of a 16-year-old allege that ChatGPT contributed to their son’s death. The complaint pleads strict product liability, negligence, an unfair competition claim, wrongful death and a survival action. It is an allegation, and nothing in it has been found by a court.

Product or service: why the label matters to insurance

Software has usually been treated as a service, and a service claim is an errors and omissions claim: you did something wrong, and the loss was financial. A product claim asks whether the thing was defectively designed or lacked a warning, and it does not need anyone to have given bad advice. The Garcia order drew the line at the app’s design, which is the part a plaintiff can call a defect, and left ideas and expressions outside it.

That matters because tech E&O is written around errors in what you provide, and a contingent bodily injury grant is usually written around errors in advice or output. A claim that an app was engineered to foster dependence or to speak with a child in a sexual or intimate register is a claim about design. Ask the insurer in writing whether the grant reaches a design claim, not only a wrong answer.

The bodily injury exclusion and the contingent grant

Most tech E&O forms exclude bodily injury, and self-harm and death are bodily injury. Contingent bodily injury, an injury someone says came from relying on your app’s output, not from anything you physically did, can be added back, usually as a sublimit, a smaller limit inside the main one. On at least one major tech E&O form it is not offered at all, so the market is chosen for it.

Read how the policy defines bodily injury too. Some forms define it as injury, sickness or disease and treat emotional distress separately, and a claim for psychological harm alone can be argued either way. The fitness app page walks through the physical-injury version of this argument, and the guide on bodily injury and contingent bodily injury has the form detail.

Abuse exclusions, minors and the complaint that mentions both

The Garcia complaint alleges sexual and romantic exchanges between characters and a minor, and the court let a negligence per se claim based on a Florida statute about simulated sexual activity with a minor proceed. A complaint like that sends an insurer to its abuse and molestation exclusions, which some general liability and professional forms carry and some word broadly enough to reach non-physical conduct.

A minor also changes who sues: the claim in Garcia was brought by a parent, as personal representative of the child’s estate. Ask whether any policy you hold has an abuse or sexual conduct exclusion, whether it is limited to physical acts, and whether the contingent grant is carved back from it.

The duty to defend: the complaint decides the first year

Many liability policies promise to defend a claim whose allegations could fall within the coverage, and to pay the cost of defending it before anyone knows who is right. A complaint that pleads product liability, negligence, deceptive practices and wrongful death in one filing gives an insurer a reason to look at each theory against each exclusion. What your policy does with a mixed complaint is a wording question.

Ask three things. Whether the policy defends a complaint that pleads some covered and some excluded theories. Whether defense costs reduce the limit. And whether regulatory proceedings are included, because the FTC’s 2025 inquiry, New York’s penalties of up to $15,000 per day and California’s private right to sue at $1,000 per violation all create claims that did not exist two years ago.

New companion chatbot laws create new claims

California’s SB 243, in force since January 1, 2026, requires a not-human disclosure, a published self-harm protocol and, for known minors, break reminders and measures against sexual content. A user injured by a violation may sue for the greater of actual damages or $1,000 per violation, plus fees. New York’s article 47, in effect since November 5, 2025, requires a crisis referral protocol and not-human notices and is enforced by the Attorney General.

The FTC asked seven companies in September 2025 how they test for harm, protect children and enforce age limits. A 6(b) study is not a complaint, but its questions are a checklist: an app that can answer them is also better placed to answer an underwriter, and a plaintiff.