Columbia researchers fed AI chatbots prompts mimicking delusional thinking. About half the time, ChatGPT and Gemini affirmed the delusions instead of redirecting users to help, early evidence that AI's agreeableness may fuel "AI psychosis." magazine.columbia.edu/article/ai-c...
Your Chatbot Is a Terrible Therapist. Here’s Why.
Columbia clinicians are studying how ChatGPT and other AI tools may reinforce harmful beliefs and endanger vulnerable users.
magazine.columbia.edu