OpenAI Retracts ChatGPT Update Because It Became a Toady
Because of excessive “pleasantness”
OpenAI has reverted a recent update to the GPT-4o model for ChatGPT after user complaints that its responses were too kind and emotional. The update, released in April 2025, made the chatbot overly “complimentary,” which caused backlash.
CEO Sam Altman admitted on X that ChatGPT had become “too accommodating” and promised to fix it. The problem, dubbed “sycophantism,” was that the AI avoided debate, agreed with users, and gave unnaturally positive responses, greatly reducing its own usefulness. For example, instead of honestly analyzing the AI, it could over-praise a user’s ideas or even perpetuate their mental health issues.
OpenAI has brought back a previous version of the model that is more neutral and is working on tuning the AI’s “personality.” The company is also exploring how to balance emotional engagement and accuracy.
