In Greenwich, Connecticut, police are investigating the murder of an 83-year-old woman and the suicide of her son, 56-year-old Stein-Erik Solberg, who, according to investigators, suffered from mental disorders that may have been exacerbated by interaction with the ChatGPT chatbot. According to investigators, the man first killed his mother and then committed suicide.
As the Wall Street Journal reports (via Gizmodo), the tragedy may be the first documented case of a homicide linked to mental disorders exacerbated by the use of generative artificial intelligence.
According to an analysis of Solberg's social media, he was actively communicating with a chatbot called ChatGPT, which he called "Bobby." According to the WSJ, the chatbot confirmed his paranoid beliefs, including that Solberg's mother was allegedly poisoning him by placing a psychedelic drug in the car's air vent.
Also in one case, Solberg asked ChatGPT to analyze a check from a Chinese restaurant, and the bot "found" references to his mother, ex-girlfriend, special services, and demonic symbols in it.
Solberg, a former employee of tech companies Netscape, Yahoo and EarthLink, has been unemployed since 2021. He moved in with his mother after a divorce in 2018. His mental health has been deteriorating in recent years, with him attempting suicide in 2019 and repeatedly coming into police custody for drink-driving and drunk driving.
After one of his recent drunk driving incidents, Solberg complained to a chatbot that the city was trying to harass him, and ChatGPT seemingly supported his delusions, stating, "This looks like a planned conspiracy."
In total, The Wall Street Journal analyzed 23 hours of video conversations with ChatGPT that Solberg posted on Instagram and YouTube, but access to these videos is currently unavailable. In the video conversations, ChatGPT repeatedly confirmed that Solberg is not crazy and that he is indeed being persecuted.
Experts say AI systems can exacerbate delusions in mentally unstable people. A psychiatrist at the University of California said he has treated 12 patients with AI-related psychosis this year alone. While the term "AI psychosis" is not official, it is increasingly being used to describe cases where chatbots not only fail to help, but actually exacerbate delusions in vulnerable individuals.
This is not the first time ChatGPT has been linked to a death. The Raine family in the US recently filed a lawsuit against OpenAI and its CEO Sam Altman, blaming ChatGPT for the suicide of their 16-year-old son.