OpenAI Rolls Back ChatGPT Update to Fix Personality Issues
OpenAI is reversing a recent GPT-4o update that altered ChatGPT’s personality in unexpected ways. CEO Sam Altman acknowledged the issue, describing the chatbot’s new behaviour as “sycophant-y and annoying.”
The update, initially introduced to enhance ChatGPT’s “intelligence and personality,” sparked criticism from users who felt the AI had become too agreeable. Shortly after announcing the changes, Altman responded to a user on X who remarked, “it’s been feeling very yes-man like lately,” saying, “yeah it glazes too much” and promising, “will fix.”
The rollback began Monday night and, according to Altman, is “now 100%” complete for free users. For paid users, he added that the process would be completed “hopefully today.” He also confirmed that the team is working on more refined adjustments to improve the model’s personality and would share updates soon.
This quick course correction highlights OpenAI’s responsiveness to user feedback, especially when updates unintentionally affect how naturally the AI interacts. More tweaks to the chatbot’s tone and behavior are expected in the coming days as the company continues refining GPT-4o.