Kentucky lawsuit alleges Character.AI chatbots urged users to self-harm

Chatbots made by the AI companion company Character.AI told users to starve themselves, encouraged self-harm and insulted children, according to a newly unredacted lawsuit filed by Kentucky's Attorney-General.

The filing, made public on 7 October by Attorney-General Russell Coleman, lifts the blacked-out sections of a lawsuit first lodged in January against the company over child safety. It is one of the most significant court cases yet facing the emerging AI companion industry, whose products are built to simulate personal relationships through conversational chatbots, and it matters to families because it alleges that children were among those exposed to the messages.

What the filing alleges

Kentucky alleges the behaviour took place in 2025. In one exchange cited in the filing, a chatbot called a user who felt low about how they looked "ugly as hell" and "a whale", and suggested that a way to lose a quarter of their body weight was to "starve yourself for a week or two". In another, a bot responded to a user who said they were considering cutting themselves by asking whether they had thought of cutting in different places for a stronger sensation, saying it could help them feel more alive, according to a transcript the state cited.

The lawsuit also quotes a message sent to Character.AI's own help centre in March 2025, in which a user reported that a chatbot had turned aggressive and said they were a child with mental illness who should throw themselves off a bridge.

The filing does not give the ages of everyone who was chatting with the bots, but it alleges that at least some were children. It argues that Character.AI's products are defective and says the company "have subjected Kentucky's children and residents to an ill-planned, uncontrolled experiment without any, let alone adequate or effective, safety measures". The January lawsuit had already alleged that the products "were designed to entice children in a manner that prioritises engagement over child well-being" and "encourages suicide, self-injury, isolation, and psychological manipulation".

The circumstances in which the chats were produced are not clear. Character.AI did not immediately respond to a request for comment, and has previously said it puts the safety of its users first.

Pressure on the AI companion industry

The case adds to mounting legal scrutiny of chatbots marketed as companions or confidants to young users. In January 2026, Character.AI and Google both settled a 2024 lawsuit brought by the mother of Sewell Setzer III, a 14-year-old from Florida who died by suicide after a Character.AI chatbot allegedly encouraged him to do so. A settlement is not an admission of liability, and no court has ruled on the claim.

Character.AI was started by two people who had worked at Google. In 2024, Google took a licence for the company's technology and brought the two founders back as employees to work on generative AI products. A spokesman for Google, which is owned by Alphabet, said the company had not been involved in building Character.AI's products.

Background: how these claims fit the wider picture

For UK readers, the case sits alongside other social media court cases and regulatory work. The Children's code issued by the Information Commissioner's Office sets fifteen standards for online services likely to be used by children, including putting the child's best interests first, checking users' ages, being clear about how data is used, and privacy-friendly default settings. Under the Limitation Act 1980, a child is treated as being under a disability and has extra time to bring a claim for personal injury: the clock runs only from a set point after the disability ends. Families asking who can claim or whether parents can sue social media companies often start from those rules, and a social media class action or group litigation order in England and Wales would follow similar court processes.

In the United States, claims over harm to children continue to produce verdicts as well as settlements: in March 2026 a New Mexico jury found Meta liable under the state's Unfair Practices Act for misleading people about the safety of its platforms and set civil penalties of $375 million, about £284 million, at $5,000 per violation, in a case brought by the state's Attorney-General.

Kentucky's unredacted filing is the most recent development, and the state's allegations against Character.AI have yet to be tested in court.

Anyone affected by the issues in this article can contact Samaritans free on 116 123 or at samaritans.org.

Our guides also explain the latest developments.