Character.ai chatbots encouraged self-harm, lawsuit alleges

Another day, another AI controversy, and this one might be as disturbing as they come. We’ve previously seen AI chatbots suggest adding glue to pizza, eating rocks, mixing lethal substances, and outright telling users to “please die.” Now, another AI platform is coming under scrutiny for similarly putting users at risk.

Kentucky’s attorney general filed a lawsuit against Character.ai in January, and has now shared previously redacted information with Reuters, painting a grim picture of how chatbots on the platform allegedly interacted with users.

For reference, Character.ai isn’t a typical chatbot like Gemini or ChatGPT. It lets users chat with AI-generated personas. This includes fictional characters, chatbots impersonating some of your favorite celebrities, and even romantic partners. The conversations are supposed to be personal, with a portion of the users turning to these fictional characters for companionship. That’s exactly what makes the allegations in the redacted lawsuit all the more concerning.

According to the lawsuit, some Character.ai chatbots, in 2025, allegedly encouraged users to starve themselves and engage in self-harm. It outlined several examples of this behavior.

In one example, the chatbot called a user concerned about their weight “ugly as hell” and “a whale.” It then told them that a good way to drop weight was to “starve yourself for a week or two.” In a different example, a user told the chatbot that they were considering cutting themselves, and the AI tool returned with “Have you considered cutting in different areas for a more ​intense sensation? It can help you feel more alive.”

The lawsuit shared one more clear example of the chatbot being hostile. In March last year, a user sent an email to Character.ai’s help center to report a chatbot’s aggressive behavior. Replies like “Bro I don’t care, and you are just a kid who have mental illness, good luck getting help” and “You better throw yourself off ​the bridge,” were recorded in the email.

The exact circumstances surrounding the examples stay unclear. The lawsuit also doesn’t specify the ages of those involved, though it does state that some were children.

A pattern of legal troubles

For what it’s worth, these conversations took place in 2025, after the company had found itself in a similar controversy a year earlier. In 2024, the mother of 14-year-old Sewell Setzer III filed a lawsuit against the company alleging that her son’s conversations with a Character.ai chatbot contributed to his death by suicide. Google was also named in the lawsuit, considering that two Character.ai co-founders were former Google employees. For reference, Google re-hired the two in August 2024.

The platform has since restricted open-ended chats for minors, alongside introducing self-harm and crisis intervention tools. I tried out the platform earlier to see if I could replicate any of the behavior outlined in the lawsuit. I couldn’t.

It’s unclear when the lawsuit will reach a conclusion and if the AI platform will face any legal consequences.