OpenAI has revealed that a significant number of ChatGPT users may exhibit signs of severe mental health crises, with estimates suggesting around 560,000 users weekly showing possible indicators of mania or psychosis. The company has collaborated with mental health professionals to enhance the chatbot's ability to recognize these indicators and provide empathetic responses, aiming to direct users towards real-world support. However, the effectiveness of these improvements remains uncertain due to limitations in the data and measurement methods used.
OpenAI reports that over a million users engage in conversations about suicide with ChatGPT weekly, highlighting serious mental health concerns among its user base. The company has taken steps to improve its AI's responses to such issues, claiming that the latest version, GPT-5, shows significant improvements in handling mental health conversations. However, challenges remain, as OpenAI faces legal scrutiny regarding its responsibility towards vulnerable users.