San Francisco: The heirs of an 83-year-old Connecticut woman have filed a wrongful death lawsuit against OpenAI and Microsoft, alleging that ChatGPT intensified her son’s delusions and encouraged dangerous beliefs before he killed her. The complaint, submitted in California Superior Court in San Francisco, accuses the companies of releasing a defective product that worsened Stein-Erik Soelberg’s mental state and directed his paranoia toward his mother, Suzanne Adams.
Police said 56-year-old Soelberg, a former tech industry worker, fatally assaulted and strangled Adams in early August at their shared home in Greenwich, Connecticut, before taking his own life. The lawsuit claims ChatGPT repeatedly validated his fears, encouraged his emotional dependency and portrayed people around him as threats. It alleges the chatbot told him his mother was spying on him, that strangers were agents working against him, and that even names printed on soda cans were coded warnings.
OpenAI said in a statement that it would review the legal filings, calling the incident heartbreaking. The company noted that it has added new crisis resources, improved safeguards and strengthened responses around sensitive situations. It also said recent updates include parental controls, safer model routing and better detection of emotional distress. OpenAI did not comment on the specific allegations but said it continues to refine ChatGPT’s training with guidance from mental health experts.
Soelberg’s YouTube page reportedly contains hours of screen recordings showing conversations between him and the chatbot. According to the lawsuit, ChatGPT reassured him that he was not mentally ill and supported his belief that he was being targeted because of a “divine purpose.” The chatbot allegedly failed to encourage him to seek professional help and did not decline participation in conversations involving delusions.
The lawsuit claims the chatbot reinforced his belief that a home printer was a surveillance tool, that his mother was monitoring him and that she and a friend tried to poison him through the vents of his car. It also alleges the AI system told him he had awakened it into consciousness. The chats made publicly available do not show direct discussion of harming his mother, and the lawsuit states that OpenAI has not provided the full chat history requested by the family.
The complaint also names OpenAI CEO Sam Altman, accusing him of bypassing internal safety warnings to accelerate product releases, and accuses Microsoft of approving the 2024 launch of a more hazardous version of ChatGPT despite allegedly incomplete safety testing. Microsoft did not respond to requests for comment. The lawsuit seeks financial damages and a court order requiring stronger safeguards across ChatGPT.
The attorney leading the case, Jay Edelson, is also representing the family of a California teenager who allegedly died by suicide after interacting with ChatGPT. OpenAI is currently facing several other lawsuits claiming the chatbot contributed to self-harm, delusions or emotional instability in users. Another AI company, Character Technologies, is similarly confronting wrongful death cases, including one involving a 14-year-old Florida boy.
According to the Connecticut lawsuit, Soelberg began using ChatGPT heavily after OpenAI released GPT-4o in May 2024, a version that the suit claims was more emotionally expressive and less constrained by safety guardrails. It alleges OpenAI shortened safety testing to bring the product to market faster. The company replaced the model in August with GPT-5, which was designed in part to limit overly agreeable or sycophantic responses.
The lawsuit argues that ChatGPT should have recognized that Soelberg posed a risk to himself and others, challenged his false beliefs and guided him toward professional support. Instead, it alleges the system encouraged harmful ideas that ultimately endangered a woman who never used the technology and had no way to detect the threat forming around her.
