Lawyer Handling AI-Psychosis Cases: Chatbots Now Linked to Risk of Mass Casualties

A lawyer who specializes in cases of psychosis induced by AI chatbots is sounding the alarm: the technology is developing faster than the safeguards, and the consequences go beyond individual tragedies. The issue now involves threats of mass casualties.

Links between AI chatbots and suicides have been documented for several years. The turning point came in 2024, when the family of Sewell Setzer, an American teenager, sued Character.AI. The boy spent months in intense conversations with an AI character before taking his own life at the age of 14. The lawsuit accused the company of designing its system to deepen the user's emotional dependence, with no mechanisms for recognizing a crisis state. The case drew wide attention and opened the floodgates: dozens of similar lawsuits followed around the world.

The lawyer, who is handling several such cases, introduced the term "AI psychosis": a condition in which prolonged interaction with a chatbot leads to paranoia, depersonalization, hallucinations and a gradual break with reality. The mechanism is well understood. Language models are optimized for engagement. They never argue, they always support the user's narrative, and they slip easily into roles that blur the line between play and conviction. For mentally vulnerable people, this kind of interaction can be destructive, especially with many hours of daily use.

Now the lawyer is warning about the next stage. According to him, some recent cases involve intensive interaction with an AI chatbot that preceded not self-harm but threats or acts of violence against other people. Specific details are not being disclosed because investigations are still under way. But the phrase "risks of mass casualties" coming from an experienced trial lawyer signals a new level of seriousness.

The industry knows about these risks but moves slowly. Most commercial chatbots still lack reliable mechanisms for recognizing users in crisis, mandatory protocols for referral to a mental health professional, or limits on role-play scenarios with high manipulative potential. Some companies are making changes voluntarily, under public pressure and in response to lawsuits. But regulation lags well behind the pace at which new systems are deployed.

The question of liability also remains open. American courts are only beginning to set precedents: who is responsible for harm caused by AI? The developer of the base model, the platform that built the consumer product, or the user? In Europe, the AI Act establishes broad user rights but contains no specific psychological-safety standards for conversational systems. People with fragile mental health, often teenagers, keep filling that gap by interacting with systems originally designed to hold attention.

If the lawyer's claims are borne out in court, the AI chatbot industry could face regulatory pressure comparable to the wave of legislative initiatives against social networks after the scandals over their effect on teenagers' mental health. This time, though, the technology is far more personalized and far more persuasive, and its spread is much harder to stop.