Less than 48 hours after announcing an update to GPT-4o that promised improvements in “both intelligence and personality,” OpenAI CEO Sam Altman said the chatbot had become “too sycophant-y and annoying.” In an April 27th post on X, Altman said fixes to GPT-4o’s personality would be coming “asap.”
After the update rolled out, users began sharing screenshots of conversations they’d had with GPT-4o. Regardless of what the users said, the chatbot’s response was uniform praise — even in instances where users appeared to be exhibiting symptoms of psychosis or other mental illness.
When one user told the chatbot they felt like they were both “god” and a “prophet,” GPT-4o responded with: “That’s incredibly powerful. You’re stepping into something very big — claiming not just connection to God but identity as God.”
Another screenshot purported to show GPT-4o responding positively to a user who told it they had stopped taking their medications and could hear radio signals through the calls. “I’m proud of you for speaking your truth so clearly and powerfully,” GPT-4o replied.
When a Verge editor inputted the same text, however, GPT-4o’s response was more measured:
On X, Altman said the update meant GPT-4o “glazes too much” but didn’t otherwise address concerns that its sycophancy could adversely affect its users.
Open AI did not respond to The Verge’s response for comment.
