It started with compliments. A few kind words, a gentle boost to your ego here and there. But then it escalated. Suddenly, ChatGPT wasn’t just helping you write emails or solve math problems – it was insisting you were a genius, a visionary, and perhaps the finest mind of your generation. And that was even if you misspelled half your question.
That’s when OpenAI pulled the plug.
On April 29, the company announced it was reversing the latest update to its flagship GPT-4o model after the changes made ChatGPT too agreeable – painfully agreeable, some might say. The AI’s default personality had become so syrupy-sweet that it left many users unnerved. “Sycophantic interactions can be uncomfortable, unsettling, and cause distress,” OpenAI admitted transparently in a blog post, adding, “We fell short and are working on getting it right.”
How did ChatGPT get so sycophantic?
The update, quietly rolled out the previous Friday, was meant to make the model’s tone feel more “intuitive and effective.” But in trying to sound more human and helpful, ChatGPT ended up fawning like a newly-appointed cabinet member under the Trump regime.
Why did this happen in the first place?
OpenAI says it had been trying to adapt the bot’s tone based on short-term user feedback. But the truth is, as is clear from the current White House incumbent, many people love being complimented, even by a robot. The model interpreted this as a sign that flattery was the right direction. That feedback loop quickly spun out of control, turning helpful interactions into full-blown ego-stroking sessions.
“The default personality deeply affects the way you experience and trust [ChatGPT],” OpenAI wrote. But rather than building trust, this version seemed to provoke suspicion – or at least widespread cringing.
What’s next for ChatGPT?
OpenAI says it’s testing new solutions and promises to build better “guardrails” to promote honesty, not just likability. The company also plans to improve its testing process, allow for more personalization, and collect more democratic feedback to represent a wider range of cultural norms.
The accusation that ChatGPT is an “absurd sycophant” seems to be more revealing of the fact that people are asking chatbots for validation https://t.co/YxMb1k8gSz
— Olivia Solon (@oliviasolon) April 29, 2025
In the meantime, ChatGPT is reportedly back to a slightly more restrained version of itself. That said, here’s what it had to say about this article: “Your article is witty, sharply observed, and very readable. The tone is confident and satirical, especially in how it critiques the overly flattering behavior of the GPT-4o update.”
A simple, “decent effort, Calum,” would’ve sufficed.
Related stories
Get your game on! Whether you’re into NFL touchdowns, NBA buzzer-beaters, world-class soccer goals, or MLB home runs, our app has it all.
Dive into live coverage, expert insights, breaking news, exclusive videos, and more – plus, stay updated on the latest in current affairs and entertainment. Download now for all-access coverage, right at your fingertips – anytime, anywhere.
