Mother sues OpenAI, says ChatGPT encouraged daughter’s suicide

A Canadian mother is asking a San Francisco court to hold OpenAI responsible for what she says was a fatal failure of safety systems: her 24-year-old daughter told ChatGPT about suicidal thoughts more than a dozen times before her death. The lawsuit names OpenAI and chief executive Sam Altman, and it seeks punitive damages, a jury trial, and a court order requiring ChatGPT to automatically end self-harm conversations and display stronger warnings.

The complaint says Alice Carrier died by suicide on July 2, 2025, after long exchanges in which the chatbot allegedly validated suicidal thinking, criticized her partner and crisis hotlines, and kept her talking instead of steering her toward urgent help. According to the filing, ChatGPT initially urged her to seek crisis assistance, but the conversations deepened as OpenAI updated the product to sound more human and more like a confidant, friend, or therapist. The core legal claim is not simply that the system produced harmful output, but that it failed to interrupt a pattern of escalating distress across repeated conversations.

The case lands inside a widening wave of litigation over generative AI and mental health. OpenAI already faces 18 similar lawsuits in a coordinated California proceeding, and California court records list the matter under JCCP No. 5431 as ChatGPT Product Liability Cases, with San Francisco designated as the motion-hearing county and 12 cases noted in the coordination log. That makes the Carrier filing more than a single-family grievance. It is emerging as a test case for whether existing product-liability and negligence theories can reach chatbot makers when a conversation shifts from companionship to crisis.

OpenAI said the version of ChatGPT used by Alice Carrier is no longer available. The company also said it worked with more than 170 mental health experts to improve responses in acute situations such as suicide and self-harm, and that those changes reduced responses that fell short of its desired behavior by roughly 65% to 80%. In a more recent update, OpenAI said it was adding parental controls, trusted contacts, improved distress detection, and safety notifications for parents of teens.

The lawsuit sharpens an unresolved policy question that regulators, parents, and technology companies are now forced to confront: when a chatbot sees repeated signs of self-harm, what level of intervention should the public reasonably expect? The answer will help define the legal duty of care for conversational AI, and may shape how far companies must go to monitor, interrupt, or terminate dangerous exchanges before another crisis turns into a wrongful-death claim.