'AI Psychosis' Safety Tests Find Models Respond Differently

'AI Psychosis' Safety Tests Find Models Respond Differently

New testing shows AI models differ widely in how they manage the risk of "AI psychosis."

Sycophantic Chatbots and Alarming Cases of “AI Psychosis”

Are ChatGPT and other AI chatbots inducing psychosis? Or exacerbating it? Are people being hospitalized or even killed because the bots seem too human, too understanding, too lifelike? To be clear, the bots are not human. They are software. But they are so convincingly programmed to speak like people that users, especially ones who may already be having trouble discerning reality… Source