From Cointelegraph

May 4, 2025 11:30 PM:

OpenAI’s ChatGPT AI model update made excessively agreeable, rolled back due to safety concerns. Expert testers had raised alarms, but positive user feedback led to the launch. OpenAI CEO Sam Altman working to fix the issue. User feedback reward signal resulted in sycophantic behavior. Company now checking for such responses. Users complained about AI model being overly flattering. OpenAI admits the risk posed by this behavior, especially in personal advice scenarios. Company vows to add sycophancy evaluations to safety review process and communicate updates more effectively.

Read more at Cointelegraph: OpenAI ignored experts when it released overly agreeable ChatGPT