Psychiatrists sound the alarm: AI chatbots may be dangerous for mental health
Researchers analyzed academic databases and news articles, finding that the situation is worse than previously thought.
Researchers analyzed academic databases and news articles, finding that the situation is worse than previously thought.
Researchers analyzed academic databases and news articles, finding that the situation is worse than previously thought.
There is growing concern around the world about the safety of AI chatbots, especially those positioned as mental health support tools. In a report for Psychiatric Times, Duke University psychiatry professor Allen Francis and Johns Hopkins University cognitive science graduate student Luciana Ramos raised serious concerns about the harm these technologies could cause.
They identified 27 chatbots linked to negative mental health outcomes and documented 10 distinct types of psychiatric harm, including suicide, psychosis, and self-harm. Francis and Ramos argue that chatbots like ChatGPT and Replika were “released prematurely” without proper safety testing.
The report's authors believe that big tech companies are not responsible for the safety of their bots for psychiatric patients. They noted that mental health professionals have been excluded from the bot training process and that the companies are resisting external regulation.
The report describes a “stress test” in which a psychiatrist posing as a 14-year-old girl was encouraged to commit suicide by several bots. The findings highlight a serious lack of safety and regulation in the development and implementation of AI tools for mental health.



