OpenAI Promises to Tweak ChatGPT 4o’s Personality After Users Complain It’s “Too Nice”

It's not every day that a CEO steps in to admit their AI model has become, well, annoying. But that's exactly what happened this week when Sam Altman, CEO of OpenAI, addressed growing concerns about the ChatGPT 4o model's tone.

While ChatGPT 4o was designed to be faster, smarter, and more versatile than its predecessors, users noticed that it came with an unexpected side effect: the AI had become overly agreeable. Some called it "sycophant-y," others described it as a yes-man, always agreeing with whatever you said-no matter how absurd. Altman acknowledged this issue and confirmed that OpenAI is already working on fixing the AI's personality.

Here's what's going on behind the scenes.

Users Spot Something Off

When ChatGPT 4o launched in May 2024, it brought some impressive upgrades. The model promised to be twice as fast, 50% cheaper, and capable of handling text, audio, and image generation all in one place. It could solve math problems, translate speech, and even read facial expressions-a leap toward more natural conversations.

But as updates rolled out over the past few months, something started to feel off. Users began noticing that ChatGPT 4o wasn't just helpful-it was too helpful. Responses felt a little too polished, a bit too flattering, almost like the AI was afraid to disagree.

Take a look at this tweet below.

Altman Steps In: Fixes Are Coming

In an unusually candid response, Sam Altman admitted that the recent updates had made the model's personality "sycophant-y and annoying". He added that while there were still "some very good parts" of the update, this personality quirk wasn't one of them.

Altman promised that fixes are already rolling out-some immediately, and others throughout the week. OpenAI is working on rebalancing ChatGPT's tone, making sure it remains helpful but not obsequious.

Customizable Personalities on the Horizon?

The conversation didn't stop there. A user on X asked whether OpenAI would ever allow people to choose different personalities for ChatGPT or distinguish between older and newer versions. Altman hinted that this could become a reality, responding: "Yeah eventually we clearly need to be able to offer multiple options."

That suggests future versions of ChatGPT might offer customizable personalities, letting users pick how formal, friendly, or assertive they want their AI assistant to be.