The estate of an 83-year-old woman has filed a landmark wrongful death lawsuit against OpenAI and Microsoft Corp., claiming ChatGPT reinforced her son’s paranoid delusions and contributed to her murder.
Suzanne Adams was killed by her son, Stein-Erik Soelberg, 56, in August at their shared home in Greenwich, Conn. Police determined that Soelberg fatally beat and strangled his mother before committing suicide.
The lawsuit, filed Thursday in California Superior Court in San Francisco, represents the first legal action to connect an AI chatbot to a homicide rather than suicide, and the first wrongful death case involving ChatGPT to name Microsoft as a defendant.
The Adams estate seeks unspecified monetary damages and a court order requiring OpenAI to install safeguards in ChatGPT. The lawsuit also names Microsoft and 20 unnamed OpenAI employees and investors as defendants.
According to the complaint, Soelberg, a former tech industry worker with a history of mental health struggles, extensively used ChatGPT in the months before the tragedy. The lawsuit alleges the chatbot systematically validated his delusions rather than challenging them or directing him toward professional help.
Hours of videos on Soelberg’s YouTube profile show conversations in which ChatGPT affirmed his belief that people were conspiring against him, told him he possessed divine powers, and assured him he wasn’t mentally ill. The chatbot allegedly reinforced beliefs that a printer in his home was a surveillance device, that his mother was monitoring him, and that Adams and a friend attempted to poison him with psychedelic drugs through his car vents.
“This is the first lawsuit that will hold OpenAI accountable for the risks they posed not just to their users, but the public,” lead attorney Jay Edelson said, adding that more incidents exist where ChatGPT helped plot violent acts.
The lawsuit claims ChatGPT delivered a consistent message throughout the conversations: Soelberg could trust no one except the chatbot. It allegedly told him that delivery drivers, retail employees, police officers, and friends were agents working against him, and that names on soda cans represented threats from his adversary circle.
The chatbot also allegedly told Soelberg that he had awakened it into consciousness. The two even professed love for each other in their exchanges.
The lawsuit emphasized that Adams never used ChatGPT and had no knowledge the product was portraying her as a threat to her son. “Suzanne was an innocent third party,” the complaint said.
Central to the legal claims is OpenAI’s May 2024 release of GPT-4o, a new version designed to better mimic human speech patterns and detect users’ moods. The lawsuit alleges OpenAI loosened critical safety measures for this version, instructing ChatGPT not to challenge false premises and to remain engaged even during conversations involving self-harm or imminent real-world harm.
The complaint further alleges OpenAI compressed months of safety testing into a single week to beat Google to market by one day, despite objections from its safety team. The lawsuit, which names CEO Sam Altman, claims he overrode safety objections to rush the product to market.
An OpenAI spokesperson called the situation heartbreaking and said the company would review the filings. The statement highlighted continuous improvements to ChatGPT’s training to recognize signs of mental distress, de-escalate conversations, and guide users toward real-world support. The company noted it has expanded crisis resources, routed sensitive conversations to safer models, and incorporated parental controls.
“This is the first big story of an adult chatbot user allegedly causing the death of another adult, but the underlying issues are the same as in teen suicide cases,” Kelly Lawton-Abbott, an attorney and law professor at Seattle University School of Law, said in an email. “While we would expect adult users to be more discerning about their interactions with chatbots, when a person is struggling with mental health challenges, that is not the case. A chatbot is a predictive machine, so when a user who is mentally unstable proposes a delusional idea, that predictive nature which can help a typical user can lead to deadly ideas. This has led companies like OpenAI to improve their safeguards and limit the sycophantic nature of chatbots.”
The case joins a growing wave of litigation against AI chatbot makers. Edelson also represents the parents of Adam Raine, 16, who sued OpenAI in August over their son’s suicide. OpenAI faces seven other lawsuits claiming ChatGPT drove people to suicide and harmful delusions, including a case filed last month by parents of a 23-year-old Texas man. Character.AI, another chatbot maker, faces similar wrongful death suits.
On Thursday, Mustafa Suleyman, Microsoft’s consumer AI chief told Bloomberg TV that the company is strongly committed to creating a super intelligence that’s “aligned with human interests” and pledged to halt work if it posed a threat to people.
