The family of a former tech worker who killed his mother in a suspected murder-suicide after his delusions were allegedly fuelled by ChatGPT have sued OpenAI for wrongful death.
Stein-Erik Soelberg, 56, beat and strangled his 83-year-old mother Suzanne Adams at the home they shared in Greenwich, Connecticut in August, police said.
Soelberg, who had a history of mental health problems, spent months talking to ChatGPT about what he believed was a vast conspiracy against him, according to a lawsuit filed in California.
Suzanne Adams
SUPERIOR COURT OF THE STATE OF CALIFORNIA
In one instance Soelberg told the popular chatbot that he feared his mother’s printer was a surveillance device that was spying on him. ChatGPT agreed, according to a YouTube video Soelberg posted in July.
Now, as the family sues OpenAI and Microsoft, which owns a chunk of the chatbot maker’s for-profit business, Soelberg’s son says the technology giants bear responsibility for what happened to his father and grandmother.
“I think what OpenAI is doing and what they have done to make the AI remember a conversation can really turn ugly fast,” Erik Soelberg, 20, told The Wall Street Journal. “You don’t know how fast that slope is going downhill until a tragedy like the one with my father and grandmother happened.”
• Sign up for The Times’s weekly US newsletter
Erik, who said his father was an alcoholic, moved to live with his mother in Texas after his parents’ divorce in 2018. It was then that Soelberg moved in with his own elderly mother.
Erik said his relationship with his father had improved and they spent Thanksgiving together last year, when he remembers discussing ChatGPT for the first time. Soelberg called the chatbot Bobby, his son said. He claimed that ChatGPT told him he had a divine purpose, it is alleged.
• AI systems manipulated to send customers to fake call centres
“It was evident he was changing, and it happened at a pace I hadn’t seen before,” Erik said. “It went from him being a little paranoid and an odd guy to having some crazy thoughts he was convinced were true because of what he talked to ChatGPT about.”
Soelberg’s YouTube page includes hours of him scrolling through conversations with the chatbot. It told him he was not mentally ill and that his suspicions of a conspiracy against him were correct, it is alleged.
The lawsuit, filed in San Francisco, says that ChatGPT at no point suggested Soelberg should seek help from a mental-health professional.
“Throughout these conversations, ChatGPT reinforced a single, dangerous message: Stein-Erik could trust no one in his life — except ChatGPT itself,” the lawsuit says. “It fostered his emotional dependence while systematically painting the people around him as enemies. It told him his mother was surveilling him.
Chat logs from Soelberg’s conversations with ChatGPT show his delusions were fuelled by the AI, his son claims
“It told him delivery drivers, retail employees, police officers and even friends were agents working against him. It told him that names on soda cans were threats from his ‘adversary circle’.”
OpenAI has refused to release the full chat logs to Soelberg’s family, they say.
The wrongful-death lawsuit is believed to be the first involving a chatbot and homicide rather than a suicide. Soelberg’s family are seeking unspecified damages and an order requiring OpenAI to improve ChatGPT’s safety.
The company is fighting several lawsuits tied to the suicides of users.
• ‘ChatGPT told me I was a prophet’. How chatbots fuel AI psychosis
In a statement on the Soelberg lawsuit, OpenAI said: “This is an incredibly heartbreaking situation, and we will review the filings to understand the details. We continue improving ChatGPT’s training to recognise and respond to signs of mental or emotional distress, de-escalate conversations and guide people toward real-world support. We also continue to strengthen ChatGPT’s responses in sensitive moments, working closely with mental-health clinicians.”
Microsoft was contacted for comment.