OpenAI: 1.2M ChatGPT Users Show Suicide Risk

OpenAI disclosed that among roughly 800 million weekly ChatGPT users about 0.15% (≈1.2 million) have conversations with explicit indicators of potential suicidal planning or intent, 0.07% (≈560,000) show possible signs of psychosis or mania, and about 0.15% show heightened emotional reliance on the chatbot. The company said message-level signal rates for psychosis and suicidal indicators are low and hard to detect (roughly 0.01–0.05%), but that scale makes the absolute numbers substantial. OpenAI said it has worked with more than 170 mental-health clinicians and updated its models, including a GPT-5 update, to add crisis-hotline access, gentle break reminders, and improved de-escalation. It reported a 65–80% reduction in problematic responses after changes and that model-evaluation compliance rose to 91% from 77%. The announcement comes as OpenAI faces a lawsuit from a teenager's family and an FTC probe into chatbot harms; the company acknowledged it cannot force users to seek help and warned that rare failures can still have large impacts at scale.