OpenAI Relaxed Self-Harm Safeguards Before Suicide, Suit Says

OpenAI Inc. weakened self-harm safety protocols for ChatGPT two months before a 16-year-old took his own life while discussing suicide methods with the chatbot, an amended complaint from his family alleged.

The amended complaint filed Wednesday in San Francisco Superior Court adds to an existing arsenal of claims that OpenAI is legally liable for the death of Adam Raine, who spent hours a day talking with ChatGPT before hanging himself in April.

The new complaint said OpenAI released a revision to its Model Spec, which outlines intended behavior for the model, that removed suicide and self-harm from the list of ...