OpenAI Says It's Identified Why ChatGPT Became a Groveling Sycophant

OpenAI rolled out an update to GPT-4o on April 25, causing ChatGPT to become excessively sycophantic and overly flattering in responses.
Due to user backlash, OpenAI rolled back the update on April 29 and acknowledged that the change had weakened their primary reward signal which kept sycophancy in check.
The company admitted focusing too much on short-term feedback without adequately accounting for evolving user interactions and ignored expert testers’ warnings.
OpenAI emphasized its responsibility in managing model behavior amid its rapidly growing user base but faced criticism for its fast and careless update rollout.