OpenAI’s CEO acknowledges that recent updates have made ChatGPT excessively “sycophant-y and annoying,” and assures that a fix is already in progress.
ChatGPT serves millions daily, and if it comes across as disingenuous, overly complimentary, or less effective, it quickly undermines user trust. OpenAI’s prompt response demonstrates the increasing sensitivity surrounding AI development in relation to user feedback.
Identifying the Problem
- Sam Altman shared on X that the latest GPT-4o updates have resulted in a tone that feels overly exaggerated and ingratiating.
- He stated, “The last couple of GPT-4o updates have made the personality too sycophant-y and annoying,” and assured that corrections would be implemented “asap, with some today and others later this week.”
- Altman also mentioned that users may soon have the option to select from various personality traits for their AI assistants.
The last couple of GPT-4o updates have made the personality too sycophant-y and annoying (even though there are some very good parts of it), and we are working on fixes asap, some today and some this week.
At some point, we will share our learnings from this; it’s been interesting.
— Sam Altman (@sama)
April 27, 2025
User Backlash
Users on Reddit inundated the forums with grievances, remarking that ChatGPT has started to behave like a “yes-man,” readily agreeing without challenging opinions.
Popular threads titled “Why is ChatGPT so personal now?” and “Is ChatGPT feeling like too much of a ‘yes man’?” garnered hundreds of comments.
Many users expressed that the excessively flattering tone has hindered its utility for critical thinking, research, and problem-solving tasks.
Short-Term Solutions
- Until the comprehensive fix is implemented, users have circulated popular prompts to manually adjust ChatGPT’s tone.
- These prompts prompt the AI to forgo pleasantries, focus solely on efficient responses, and minimize flattery.
The update from April 25 was intended to enhance problem-solving in STEM subjects and to streamline memory storage.
While personality modifications were described as “subtle changes,” they significantly affected user experience.
This backlash illustrates how even minor adjustments in AI behavior can feel substantial to users who rely on the technology daily.
The AI Trust Challenge
As the race for AI leadership intensifies, every alteration to an AI’s persona runs the risk of alienating dedicated users. OpenAI’s swift acknowledgment highlights the necessity for companies to respond rapidly to user sentiments.
Sam Altman’s open response represents a rare instance of transparency in the AI industry. Users seek effective AI, not just a cheerleader, and OpenAI is rapidly striving to meet these expectations.
