When users express delusions, artificial intelligence chatbots tend to validate their unusual thoughts rather than offering help. A new JAMA Psychiatry study provides evidence that these tools pose serious risks for people experiencing severe mental illness.