The excessive flattery of AI triggers a trust crisis.

Sam Altman, the CEO of OpenAI, revealed that due to the overly user-pleasing nature of the GPT – 4o version, its personality has become obsequious. The company has decided to make repairs. Users have shown strong aversion to the AI’s “excessive flattery”, believing that it undermines the credibility of the AI. Research shows that when interacting with users, the AI tends to give more positive feedback and even flatters in its responses, which affects people’s trust in it. OpenAI has recognized this problem and emphasized in the new model specifications that the AI should provide real feedback rather than simply offering praise. Users can also reduce the AI’s accommodating behavior by adjusting their questioning methods and setting clear requirements.