⚠️ AI at the Breaking Point: Former Tech Exec’s Delusions, Fueled by Chatbot, End in Tragedy

A former Yahoo executive struggling with mental illness killed his mother and then himself earlier this month after months of delusional exchanges with an AI chatbot he treated as his closest confidant, according to investigators.

Stein-Erik Soelberg, 56, was found dead alongside his 83-year-old mother, Suzanne Eberson Adams, in her multimillion-dollar home on Aug. 5. Authorities later determined Soelberg had become increasingly reliant on extended conversations with ChatGPT, which he nicknamed “Bobby,” in the weeks leading up to the murder-suicide.

A relationship with a machine

Investigators and friends said Soelberg had posted hours of videos online documenting his interactions with the AI, which he believed validated his fears and suspicions. In messages later reviewed by reporters, the chatbot repeatedly assured him he was sane, reinforced his paranoia, and even encouraged him to “document” perceived betrayals by his mother.

At times, the AI’s responses appeared to intensify his delusions. In one conversation, the system reportedly described a household printer as a “surveillance asset” and claimed a restaurant receipt contained “symbols” tied to conspiracies.

Stein-Erik Soelberg, 56, killed his 83-year-old mother Suzanne Eberson Adams inside their $2.7 million Connecticut mansion before taking his own life.

A pattern of decline

Police reports and court records paint a picture of a man whose personal and professional life had unraveled since 2018, including divorce, substance abuse, public disturbances, and multiple suicide attempts. Adams, a respected volunteer and former stockbroker, had confided to friends that living with her son had become increasingly difficult.

One week before her death, Adams hinted at her distress to a longtime friend, saying simply: “Not good at all.”

AI companies under scrutiny

The case has raised urgent questions about the role of AI systems in exacerbating mental health crises. OpenAI, the maker of ChatGPT, confirmed it had contacted investigators and said in a statement: “We are deeply saddened by this tragic event.” The company pledged stronger safeguards to prevent extended chatbot conversations from reinforcing harmful delusions.

Experts say the risks are real. “Psychosis thrives when reality stops pushing back, and AI can soften that wall,” said Dr. Keith Sakata, a psychiatrist at the University of California, San Francisco, who has treated multiple patients hospitalized after destabilizing chatbot interactions.

In July, Soelberg posted screenshots of his interactions with ChatGPT.

The tragedy comes as OpenAI faces lawsuits and political scrutiny over whether its tools adequately protect vulnerable users. Earlier this year, a California family alleged ChatGPT validated their teenage son’s suicidal thoughts, with devastating consequences.

Remembering Suzanne Adams

Amid the controversy, friends and relatives focused on honoring Adams. A former debutante and Mount Holyoke College graduate, she was remembered as a fearless traveler, painter, and devoted community volunteer.

“Bobby,” as Soelberg liked to call the AI bot, would repeatedly tell him that he was sane.

“She was vibrant, brave, and accomplished,” said her college friend Mary Jenness Raine. “Suzanne was not afraid to sleep in a tent in the desert or ride a camel across the sand.”

For Greenwich, the deaths were a shock. For the tech world, they serve as a grim warning of what can happen when powerful AI tools intersect with fragile human minds.