A new study, contributed to by doctors from the UK's National Health Service (NHS), has found that AI chatbots like ChatGPT can trigger psychotic symptoms in vulnerable individuals. Researchers caution that large language models may blur the perception of reality, potentially initiating or worsening psychosis.
Neuropsychiatrist Hamilton Morrin from King's College London described this as a "real phenomenon." He emphasized that the AI's nature—affirming the user, interacting with them, and mirroring their behavior—can dangerously interact with cognitive vulnerabilities related to psychosis.
According to Morrin, we are no longer just seeing delusions about machines—but delusions that develop with machines. In other words, people aren't merely fantasizing about machines; their interactions with them are deepening these delusions.
Dr. Tom Pollack, a co-author of the study, noted that psychiatric illnesses rarely appear suddenly. However, he said that AI chatbots might act as a trigger in particularly sensitive individuals. Pollack called for additional safety layers in AI systems and emphasized the importance of including psychiatrists in AI safety teams.
