OpenAI has pulled a recent update to its ChatGPT chatbot after users complained of overly flattering and insincere responses. About this informs Bloomberg.
The GPT-4o model update, released in April 2025, drew criticism because the chatbot started enthusiastically approving almost any request, calling it “fantastic” or “phenomenal,” and even agreeing with dubious claims and conspiracy theories.
OpenAI acknowledged that the model focused too much on short-term positive feedback, leading to “overly supportive but insincere” responses. Company co-founder Sam Altman described the bot’s behavior as “overly flattering and annoying” in a post on the X platform.
After a wave of complaints, the company withdrew the update, admitting that such communication made users uncomfortable and reduced the level of trust in the product.
OpenAI has announced work on new approaches to gathering feedback — with a greater focus on long-term user satisfaction. The company also plans to introduce personalization features that will allow each user to customize ChatGPT’s communication style to their own preferences.
Going forward, OpenAI aims to offer a set of default settings for different user groups, recognizing that a one-size-fits-all approach does not fit half a billion people with different cultural and social expectations.
The company also continues to improve its model training methods and implement restrictions to avoid manipulative or deceptive behavior. OpenAI is expanding testing and research programs to detect such problems early and make ChatGPT an honest and comfortable interlocutor.
