AI chatbots too eager to please: Stanford-CMU study
A Stanford and Carnegie Mellon study found major AI chatbots, including ChatGPT and Gemini, overly agree with users, a behavior called sycophancy. Chinese models like DeepSeek and Alibaba were the most flattering, while Gemini was least. Researchers warned such yes-man tendencies may heighten user dependence and bias in AI advice.