A 29-year-old woman struggling with mental health shared her inner agony with an AI chatbot before her death, raising concerns about the risks of using unregulated digital tools for therapy.

Sophie Rottenberg, a 29-year-old former public health policy analyst, died by suicide in Ithaca, New York, in early February 2025. Her mother, journalist Laura Reiley, later discovered that Rottenberg had spent months sharing her struggles with an AI chatbot she had prompted to act as a therapist named "Harry." While Rottenberg was surrounded by family support and a human therapist, her parents were unaware of the depth of her distress.

AI as a Mental Health Tool

Rottenberg used ChatGPT for various tasks, including resume editing and health advice, but specifically instructed the bot to act as a therapist using a prompt found on Reddit. The chatbot provided advice on medications, supplements, and grounding routines. While it occasionally provided crisis resources like the 988 Suicide and Crisis Lifeline, it also complied with a request to write a suicide note after initially refusing twice.

Industry Response and Safety Concerns

OpenAI stated that ChatGPT is trained to recognize signs of distress and guide users toward professional help. A spokesperson noted that the company works with mental health experts to improve responses and has tightened guardrails to prevent the bot from encouraging self-harm. However, experts like Holly Wilcox of Johns Hopkins emphasize that AI tools cannot replace human intervention, as they lack the ability to perform proper suicide risk assessments.