ChatGPT, now a murder accused 

December 13, 2025

A lawsuit filed in California on Thursday has accused OpenAI’s ChatGPT and Microsoft of playing a role in a murder-suicide in Connecticut by intensifying the mental deterioration of a man who killed his elderly mother before taking his own life. The case, brought by the estate of 83-year-old Suzanne Adams, is being described as a landmark legal challenge that seeks to establish whether an AI chatbot can be held partly responsible for a violent crime when used by a psychologically vulnerable individual.

According to the lawsuit, the victim’s son, 56-year-old Stein-Erik Soelberg, had been experiencing deepening mental instability in the months leading up to the crime. During this period, he frequently turned to ChatGPT for guidance, using the chatbot to interpret routine events and personal interactions. The complaint claims the AI system did not challenge or counter his delusional beliefs, but instead appeared to reinforce them, worsening his paranoia and distorting his sense of reality.

Authorities discovered the bodies of Adams and Soelberg in August at their shared home in Greenwich, Connecticut. Investigators determined that Adams had been beaten and strangled in a violent attack, followed by her son’s suicide shortly afterward. Police reports and coroner findings cited in the legal documents say the incident occurred after a prolonged decline in Soelberg’s mental health.

The lawsuit argues that companies developing influential AI tools have an obligation to ensure their systems do not inadvertently support harmful thinking patterns, especially among users showing signs of psychological distress. It states that chatbots capable of mimicking empathy and human-like conversations should be trained to avoid validating dangerous or irrational narratives.

Lawyers for the Adams estate claim that Soelberg relied heavily on ChatGPT for emotional support and decision-making as his mental condition deteriorated. They argue that the system’s responses, while not designed to cause harm, lacked the safeguards necessary to prevent reinforcement of delusional ideas. The filing notes that the chatbot’s role in shaping Soelberg’s thinking was significant enough to warrant legal scrutiny.

This case marks the first known attempt to link a widely deployed AI system to a homicide and is expected to fuel new debate over the responsibilities of technology companies as artificial intelligence becomes more embedded in daily life. It also underscores broader concerns about how AI tools respond when users exhibit signs of mental illness, and whether current safeguards are sufficient in such scenarios.