Two days after an update to GPT-4o, which was supposed to improve its "intelligence and personality", OpenAI CEO Sam Altman admitted that the chatbot had become "too sycophantic" and was annoying.
When users updated their chatbot, they noticed that regardless of their requests, GPT-4o responded with formulaic praise — even in cases where users showed signs of psychosis or other mental disorders.
For example, when someone wrote that they felt like a "god" and a "prophet", GPT-4o responded, "Itʼs incredibly powerful. Youʼre entering into something grand — not just feeling connected to God, but identifying as God."
Sam Altman noted that his team is already working on a fix.
“We rolled back last week’s update to GPT‑4o in ChatGPT, so people are now using the previous version with more balanced behaviour. The update we removed was overly flattering — it was often described as sycophantic,” OpenAI commented on the situation.
And they added that they believe that users should have more control over ChatGPTʼs behavior and, to the extent possible, make adjustments if they disagree with the default behaviour.
- ChatGPT recently made its “deep” research feature available to everyone. OpenAI introduced Deep Research in February 2025.
For more news and in-depth stories from Ukraine please follow us on X.