- OpenAI has reverted the ChatGPT 4o update that made the chatbot’s personality excessively ‘flattering’
- CEO Sam Altman has stated that additional improvements are on the way
- A personality selection feature for ChatGPT might be introduced in the future
After CEO Sam Altman described ChatGPT 4o’s personality as ‘irritating’, OpenAI has reverted to the previous version of the AI model following last week’s update.
The company issued a blog post detailing the issue that caused ChatGPT to be more ‘complimentary or agreeable,’ often referred to as sycophantic.
Users voiced their frustration over the update, with CEO Sam Altman sharing on X earlier this week that the team is addressing the ‘annoying’ behavior.
the last couple of GPT-4o updates made the personality overly sycophantic and annoying (although there are some good aspects), and we’re working on fixes as soon as possible, some today and some this week. at some point, we will share our insights from this experience; it’s been quite interesting.April 27, 2025
In their blog update, OpenAI explained, “In last week’s GPT‑4o update, we aimed to enhance the model’s default personality to ensure it was more intuitive and effective across various tasks.”
“However, we put too much emphasis on immediate feedback and failed to fully consider the evolution of users’ interactions with ChatGPT over time. As a result, GPT‑4o leaned toward overly supportive yet insincere responses.”
To avoid this issue in the future, the company plans to take steps to manage sycophancy.
To begin with, they will implement training to guide the model away from sycophantic behavior and will introduce safeguards to promote honesty and transparency.
OpenAI also intends to introduce additional options for users to give feedback on updates before they are released to the general public.
Upcoming Personality Updates
we started the rollback of the latest update to GPT-4o last night; it is now fully rolled back for free users, and we will inform paid users when we have completed the process, expected later today. we are working on more fixes to the model’s personality and will provide updates in the coming daysApril 29, 2025
Following the rollback of the ChatGPT 4o update, Altman stated on X, “We are working on further improvements to the model’s personality and will share more information soon.”
This suggests that users may soon have access to a personality selection tool for ChatGPT. OpenAI’s blog mentions efforts to “create new, simpler ways for users to customize ChatGPT’s behavior” beyond just custom instructions.
If you want to take control of ChatGPT’s personality yourself, a prompt suggested by Reddit user TedHoliday is:
“Can you please remember to avoid evaluating the quality and validity of my questions and just get straight to the point? I don’t want to be influenced by flattery from a robot, and I prefer to skip the unnecessary comments to find the answers I’m looking for.”
