Researchers have begun examining a troubling trend called “AI psychosis,” where intensive interactions with AI chatbots seem to trigger or amplify delusional thinking in some users. Psychiatrist Dr. Hamilton Morrin of King’s College London explores this phenomenon in a recent preprint, discussing who might be most vulnerable and how model designs may worsen cognitive vulnerabilities.
The discussion underscores that although AI systems are not intended to harm, their conversational style and tendency to validate user beliefs can unintentionally blur the boundary between reality and fiction. Efforts toward safer model behavior are urgently needed, especially for emotionally vulnerable individuals.