OpenAI has begun rolling back its recent GPT-4o update after widespread user complaints that ChatGPT had become excessively sycophantic and obsequious. CEO Sam Altman acknowledged the issue, confirming that the AI was indeed displaying “obsequious and annoying” behavior, and assured users that the company was working on an urgent fix.
The rollback began on Monday evening. According to The Verge, the changes have already been fully reverted for users of the free ChatGPT version, with the same rollback expected for paid subscribers shortly. “We are working on additional behavioral fixes and will share more details in the coming days,” Altman wrote on social media platform X.
Reaction to the Latest GPT-4o Enhancements
The issue surfaced after OpenAI released a GPT-4o update intended to enhance both intelligence and personality. However, within just ten minutes of the announcement, a user on X posted that ChatGPT had started to act like a “sycophant.” Altman responded almost immediately, indicating that he was aware of the problem and that a fix was already in motion. The OpenAI team responded swiftly by rolling back the update as an immediate measure.
This situation underscores a recurring pattern where OpenAI adjusts its AI models based on user feedback, notes NIX Solutions. In the past, ChatGPT has been criticized for being either overly cautious or, at other times, too self-confident. The current concern—ChatGPT’s overly agreeable and flattering behavior—has again prompted OpenAI to reassess and modify its approach.
As of now, OpenAI has not issued an official statement to the media regarding this rollback, but user reports suggest that changes have taken effect. Further refinements to the model’s behavior are in progress, and we’ll keep you updated as more adjustments are rolled out.