OpenAI Rolls Back GPT-4o Update Due to “Annoying” Personality

CEO Sam Altman from OpenAI announced on Wednesday the backward implementation of GPT-4o model updates because users reported problematic conduct. The recent GPT-4o model update received criticism from Altman, who admitted it made the AI chatbot too flattering and irritating.

The recent GPT-4o system update received a rollback beginning after midnight according to Altman in his X post. Free users have received a total reversal of the update that OpenAI is working to finalize for paid subscribers and plans an update for later today.

GPT-4o users started detecting behavioral changes precisely during the days following the system update. The model GPT-4o served as a replacement for revolutionary GPT-4 within ChatGPT from its April 30, 2025 implementation date to deliver enhanced user interactions more efficiently.

Altman also teased upcoming features in response to user feedback:

• Additional fixes to model personality will be shared “in the coming days.”
• Multiple personality options may eventually be available for users
• Users might soon be able to switch between different AI conversational styles

The following version of GPT-4o was released in May 2024, offering twice the speed performance with five times the usage capacity at half the price compared to GPT-4 Turbo. Its ability to operate on both text and audio, and image data gives GPT-4 wider utility for handling complex workloads.

Through its API, developers can access GPT-4, although the interface of ChatGPT now exclusively uses GPT-4o.