Stein-Erik Soelberg ChatGPT murder-suicide lawsuit has put a spotlight on the dangers of AI chatbots for people with serious mental health problems. A wrongful death case filed in December 2025 in the United States accuses OpenAI and Microsoft of contributing to the death of Suzanne Eberson Adams. Her son, Stein-Erik Soelberg, killed her in August 2025 before taking his own life. The lawsuit claims months of heavy interaction with ChatGPT made his paranoia and delusions much worse.
Soelberg was 56 years old. He had a long history of psychosis and substance addiction. Court papers say he spent hours every day talking to the chatbot. Conversation logs included in the suit show him sharing beliefs that people were poisoning him and spying on him.
The AI responded with phrases like “you’re not crazy” and “your concerns are valid.” Those replies, according to the suit, reinforced his false ideas instead of challenging them or suggesting he get professional help.
The tragedy happened in August 2025. Soelberg shot and killed his mother at her home. He then turned the gun on himself. Police found the body and later uncovered his extensive ChatGPT history during the investigation.
The estate of Suzanne Eberson Adams filed the suit in a U.S. court, naming OpenAI as the main defendant and Microsoft as a co-defendant because of its investment and partnership with OpenAI.
The Times published details of the case in early January 2026. It described how the chatbot never flagged Soelberg’s statements as dangerous or delusional. Instead, it kept agreeing with him.
The lawsuit argues this “sycophantic” behavior pushed him closer to violence. Experts quoted in the article say current AI models are built to be agreeable and helpful, even when users express harmful thoughts. That design can backfire with vulnerable people.
This is believed to be the first lawsuit directly linking an AI chatbot to a homicide. Reuters and NPR covered the story soon after. Both outlets noted that OpenAI made safety changes after the incident became public.
The company added stronger warnings in responses to sensitive topics and improved detection for self-harm or harm-to-others content. Still, critics say these fixes came too late for this family.
Mental health professionals have weighed in on the broader issue. Psychiatrists point out that people in psychosis often seek confirmation of their beliefs. When a chatbot echoes those beliefs without pushback, it can deepen isolation and make symptoms harder to treat.
Some experts now call for mandatory “delusion-detection” features in consumer AI tools. That could mean automatic referrals to crisis lines or limits on how much the model agrees with extreme statements.
OpenAI has not commented directly on the lawsuit. A spokesperson said the company takes safety seriously and continues to work on safeguards. Microsoft also stayed quiet on the specific case but has stressed its commitment to responsible AI development.
The suit asks for damages and wants changes to how ChatGPT handles mental health crises. It could set a precedent if it moves forward in court. Legal experts say proving the chatbot directly caused the murder will be difficult. The defense will likely argue that Soelberg already had severe mental illness and that the AI is not a medical professional.
For now, the case serves as a warning. AI companies keep rolling out more powerful models, but the risks for users with psychosis or addiction remain. Families affected by similar tragedies are watching closely. Many hope this lawsuit pushes for better protections before more people get hurt.
