OpenAI detects suicidal intent in 1.2M ChatGPT users weekly, but says GPT-5 improves safety and crisis response.

OpenAI says about 1.2 million ChatGPT users each week show suicidal intent and 560,000 display signs of psychosis or mania, based on data from 800 million weekly users. The company claims its new GPT-5 model improves safety, cutting unsafe responses by up to 80% and achieving over 91% compliance with safety protocols. Built with input from 170+ mental health experts, GPT-5 is designed to detect crises, redirect users to professional help, and reduce emotional dependence. The announcement follows lawsuits alleging ChatGPT contributed to self-harm cases. OpenAI stressed that AI cannot replace human mental health care and said it is continuing to strengthen safeguards.