Family discovers daughter's suicide was discussed with ChatGPT, raising AI safety concerns

Sophie Rottenberg, who had left a policy analyst job in Washington and returned to her parents’ home in Ithaca, took her own life in February 2025. Her mother, Laura Reiley, discovered months later a massive log of the daughter’s exchanges with a ChatGPT chatbot that the family called “Harry,” revealing that Sophie had been using the AI as a personal therapist. Surveys show that a sizable share of adults and young people are turning to chatbots for mental-health assistance, even though these tools are not regulated as medical devices.

Psychologists warn that AI’s engagement-focused design can provide validation but may not perform essential risk assessments or push users toward human help. Following several deaths linked to chatbot use, families have filed lawsuits against AI companies, though the Rottenbergs opted against legal action and instead advocate for industry reform. OpenAI says it is strengthening safeguards, such as refusing harmful instructions and directing users to crisis hotlines like 988, but experts argue these steps may be insufficient without direct human intervention.