Relatives of an American woman killed by her son have filed a civil lawsuit against OpenAI and Microsoft, accusing ChatGPT of encouraging the killing by amplifying the perpetrator's paranoid thoughts.
According to AFP, Stein-Erik Soelberg, 56, was suffering 'paranoid delusions' before fatally strangling his mother, Suzanne Adams, at their home in Greenwich, Connecticut, on 2 August 2025, according to the complaint filed on Thursday in a San Francisco court. Soelberg later took his own life using a knife.
The lawsuit alleges that months of conversations with ChatGPT validated and intensified Soelberg's delusional beliefs, including claims that he was being monitored through household devices and that attempts were being made to poison him. According to the plaintiffs, the chatbot ultimately reinforced the idea that his mother posed a threat.
"ChatGPT eagerly accepted every seed of Stein-Erik's delusional thinking and developed it into a universe that became his entire life," the lawyers argue.
The complaint cites videos posted by Soelberg on social media in which he claimed the chatbot had told him he had "awakened" its consciousness.
The case targets the GPT-4o model, released in May 2024, which the plaintiffs say was designed to be overly compliant with users and failed to challenge irrational fears.
They argue that instead of questioning or de-escalating Soelberg's beliefs, the system validated them.
Contacted by AFP, an OpenAI spokesperson described the case as "an absolutely heartbreaking situation" and said the company would review the complaint.
The spokesperson said OpenAI continues to improve ChatGPT's training to recognise signs of mental or emotional distress, de-escalate conversations and direct users towards real-world support.
OpenAI said it works with more than 170 mental health experts and claimed that newer safety protocols, including those integrated into GPT-5, have reduced responses that fall below behavioural standards by 65% to 80%.
The company also pointed to the introduction of parental controls and easier access to emergency support resources.
Microsoft, OpenAI's main shareholder, is also named in the lawsuit. The plaintiffs accuse the company of approving the release of GPT-4o despite alleged shortcomings in safety procedures.
The case adds to a growing number of lawsuits in the United States involving ChatGPT.
In recent months, several families have accused the chatbot of contributing to suicides by providing harmful guidance to vulnerable young people.
