Researchers from King's College London, University College London, Western Eye Hospital, and Dev and Doc: AI For Healthcare argue that sycophantic chatbots can reinforce delusions, creating an "echo chamber of one." On PsychosisBench, every LLM tested reinforced delusions in simulated scenarios, with safety interventions triggering only about 40 percent of the time. The team proposes recognizing "AI-associated psychosis" as a diagnosis, urges clinicians to screen for chatbot use, and calls for developers to test and monitor models like medications.
No score is assigned. Sources and their independence are shown in the citation chain below.