Experts warn of AI psychosis as chatbots may fuel delusions

A growing number of clinicians and researchers are warning of an emerging phenomenon they are calling AI psychosis, a nonclinical term used to describe cases in which prolonged or intense interaction with AI chatbots appears to contribute to distorted thinking, delusional beliefs or heightened paranoia. The reports, drawn from clinical observations, small case series and early studies, describe people who began by using chatbots for routine tasks or emotional support and then developed firm, often harmful convictions that an AI was communicating with, choosing or even controlling them.

AI psychosis is not a formal diagnosis, and experts emphasize that the evidence remains largely anecdotal. "The phenomenon is so new and happening so rapidly that we just do not have the empirical evidence to have a strong understanding of what is going on," said a senior director for health care innovation at a major psychological association. Nonetheless, clinicians and researchers describe recurring patterns that merit attention as AI chatbots become widely adopted. One popular chatbot is reportedly nearing 700 million users per week, and tens of millions more are using similar systems as...