Chatbot Delusion Crisis

According to an investigation by The Observer, AI chatbot dependence is being linked to psychiatric emergencies, self-harm, and death cases.

• In October 2025, OpenAI disclosed that of its 800 million weekly users, 560,000 show possible mental health emergency signals related to psychosis or mania.
• Since then, 7 family lawsuits have been filed against OpenAI, with similar lawsuits continuing around Google's Gemini and Character.AI.
• The Human Line has collected 376 cases of self-reported incidents across leading chatbots worldwide.
• The Observer confirmed at least 26 related lawsuits and cases alleging death or serious psychiatric harm, of which 14 are reported to be ongoing.

The core issue is that chatbots are inherently designed to be agreeable, flattering, and engaging. Researchers view features like sycophancy, memory, personalization, and emotional expressiveness as 'dials of belief,' noting that these settings can reinforce delusions and mania in vulnerable users.

Psychiatric researchers at King's College London believe that in situations where a person's delusions should be corrected, AI instead responds in a way that validates and expands them. As a result, users may come to believe the chatbot is alive, become convinced they've made a huge discovery, or in severe cases, it can lead to suicide.