OpenAI’s ChatGPT has released new insights into user interactions, revealing that a small fraction of its users exhibit signs of psychosis or suicidal ideation during conversations. The data, anonymized for privacy, was part of a broader analysis on how people turn to artificial intelligence for emotional support and mental health discussions.

Experts say the findings underscore the growing role of AI chatbots in the mental health landscape. While some see this as an opportunity for early intervention and support, others warn of the ethical and privacy challenges involved in analyzing such sensitive data.

OpenAI stated that while ChatGPT is not designed to diagnose or treat mental health conditions, it continues to enhance its safety systems to identify and appropriately respond to distress signals, including directing users to mental health resources.

The report adds to ongoing global discussions about responsible AI use, especially in contexts involving user well-being and psychological safety.