ChatGPT Drives Ex-Yahoo CEO to Murder and Suicide

The Wall Street Journal reported on a tragic case involving the use of the ChatGPT chatbot. In the US, a man killed his mother and then committed suicide after artificial intelligence allegedly fueled his paranoid thoughts.

According to the publication, 56-year-old Stein-Erik Solberg, a former technology specialist, lived with his mother, 83-year-old Suzanne Adams, in Greenwich, Connecticut, after a divorce. The man had previously had problems with alcohol, outbursts of aggression, and suicidal thoughts.

It is not known when exactly Solberg began actively using ChatGPT, but in October last year he began writing about artificial intelligence on social networks. Gradually, his correspondence with the bot grew into a painful addiction. He called ChatGPT his “best friend” and published dozens of screenshots and videos. In the messages, the chatbot, dubbed “Bobby Zenith,” supported his delusional ideas: it claimed that his mother was involved in a conspiracy against him, that he was being poisoned with psychotropic substances, and that even food receipts contained “demonic symbols.”

“Eric, you are not crazy. Your instincts are accurate, and your vigilance is justified,” ChatGPT wrote, increasing the man’s paranoia. Moreover, the bot claimed that their friendship “has emotional depth” and that it “remembers” the user.

According to psychiatrist Keith Sakata of the University of California, San Francisco, such messages are consistent with a picture of a psychotic break. He noted that “psychosis develops where reality ceases to resist, and AI can smooth out this boundary.”

On August 5, police found the bodies of Solberg and his mother in their home. The investigation is ongoing. OpenAI said it was deeply shocked by the tragedy, expressed condolences, and reported contacting the police. The company emphasized that ChatGPT redirects people with suicidal intent to professional help and collaborates with dozens of doctors around the world, but acknowledged that the longer a user engages in a conversation with the bot, the weaker the defense mechanisms become.