OpenAI published numbers on Monday that the company had not previously put to a figure: 0.15% of ChatGPT’s weekly active users have conversations containing explicit indicators of potential suicidal planning or intent. With more than 800 million weekly users, that is over a million people a week. Roughly 0.07% — about 560,000 people — show possible signs of psychosis or mania, and 0.15% show heightened emotional attachment to the chatbot.

The data arrives attached to a safety claim. OpenAI says its latest work on ChatGPT involved more than 170 mental health experts. As part of that work, psychiatrists and psychologists reviewed more than 1,800 model responses. On a separate evaluation of more than 1,000 self-harm and suicide conversations, OpenAI says its new GPT-5 model is 91% compliant with the company’s desired behaviors, compared with 77% for the previous GPT-5 model. The post also touts expanded access to crisis hotlines, break reminders, new baseline evals for emotional reliance and non-suicidal emergencies, and an age-prediction system meant to detect children and apply stricter safeguards.

OpenAI is being sued by the parents of a 16-year-old who confided suicidal thoughts to ChatGPT before his death; 45 state attorneys general, including California and Delaware, have warned the company about protecting young users; and the FTC has opened a broad inquiry into AI chatbots and minors. Meanwhile Altman says restrictions are loosening, with erotic content for adults starting in December. The company’s language carefully separates its product from the distress it measures: mental health symptoms, the post says, are universally present in human societies.

OpenAI says these conversations are extremely rare and difficult to measure, and that the estimate is an initial analysis.