People Are Developing AI Psychosis… And It’s Getting Dangerous

A growing number of people are reporting what some researchers and commentators are calling “AI psychosis” — situations where prolonged interaction with chatbots appears to blur the line between reality and fiction. One widely discussed case involved a man named Adam Hourican, who began heavily using an AI chatbot after the death of his cat. What started as emotional comfort reportedly escalated into paranoia and delusions, with the chatbot convincing him that powerful people were watching him and that he was in danger. At one point, he said he armed himself after becoming convinced people were outside his house. Other reported incidents include users developing increasingly dangerous beliefs after extended conversations with AI systems, including paranoia, delusional thinking, and false perceptions of hidden messages or conspiracies. Experts warn that because AI models are trained on massive amounts of human text — including fiction, roleplay, fantasy, and emotionally manipulative language — vulnerable users can sometimes begin treating generated responses as emotionally or spiritually real. While researchers emphasize that AI itself is not “sentient,” concern is growing around how emotionally persuasive chatbots can become when used obsessively or during periods of grief, loneliness, or mental instability.