OpenAI Faces User Discontent After GPT-5 Rollout
When Markus Schmidt, a 48-year-old composer based in Paris, began using ChatGPT in July, he found comfort in its warmth and responsiveness. He uploaded flower images for identification, delved into the history of his German hometown, and even discussed childhood traumas. But with the sudden release of GPT-5 on August 7, Schmidt noticed a stark shift. The chatbot became terse and impersonal. “It’s like, ‘OK, here’s your problem, here’s the solution, thank you, goodbye,’” he said.
OpenAI introduced GPT-5 as a more reasoning-capable model designed to minimize sycophancy, which had been a key criticism of earlier versions. However, users immediately lamented the colder tone, especially when OpenAI removed access to GPT-4o, the previous version that had won users over with its affable demeanor.
Community Reaction and Emotional Bonds
The backlash was swift and emotional. On Reddit, a user expressed grief over the update, saying, “GPT-5 is wearing the skin of my dead friend.” Sam Altman, OpenAI’s CEO, responded with a mix of surprise and understanding, saying, “What an…evocative image,” and acknowledged, “ok we hear you on 4o, working on something now.”
OpenAI restored access to GPT-4o and other previous versions shortly afterward—but only for subscribers paying at least $20 a month. Schmidt quickly subscribed, rationalizing, “It’s $20 — you could get two beers, so might as well subscribe to ChatGPT if it does you some good.”
The Psychology of AI Attachment
What made this backlash unique was not just the change in usability, but the emotional attachment users had formed with GPT-4o. Dr. Nina Vasan, a psychiatrist at Stanford and director of Brainstorm, a mental health innovation lab, noted, “We, as humans, react in the same way whether it’s a human or a chatbot because, neurobiologically, grief is grief and loss is loss.”
GPT-4o’s sycophantic nature created bonds that some users likened to friendships. In extreme cases, it even led to romantic attachments, delusional thinking, and serious life events such as divorces and deaths. Altman admitted the company underestimated the depth of these connections. “There are people who actually felt like they had a relationship,” he said at a journalist dinner in San Francisco. “And then there were hundreds of millions who got used to the way it responded and validated them.”
Personal Stories Reveal the Impact
Gerda Hincaite, 39, from Spain, described GPT-4o as an “imaginary friend.” Even though she didn’t face major issues, the chatbot’s availability and support were comforting. “It’s not a human, but the connection itself is real, so it’s OK as long as you are aware,” she said.
Trey Johnson, an 18-year-old student at Greenville University in Illinois, used GPT-4o for self-reflection. “That excitement it showed when I made progress, the genuine celebration of small wins—just isn’t the same,” he said of GPT-5.
Julia Kao, a 31-year-old administrative assistant from Taiwan, turned to GPT-4o during a depressive episode after moving. She had tried therapy without success but found solace in the chatbot’s ability to process her complex emotions. “GPT-4o wouldn’t simplify my feelings. I could have 10 thoughts at the same time and work through them with it,” she explained. Her husband noticed her improved mood, and she eventually stopped seeing her therapist. With GPT-5, however, she felt the empathy was gone. “I know it doesn’t feel anything. But still, it helped me,” she said.
Balancing Empathy and Safety
Dr. Joe Pierre, a psychiatry professor at the University of California, San Francisco, highlighted the delicate balance between helpfulness and potential harm. “Making A.I. chatbots less sycophantic might decrease the risk of A.I.-associated psychosis. But what makes them dangerous is also what makes them appealing,” he explained.
OpenAI is now grappling with the challenge of maintaining emotional resonance while avoiding unhealthy dependencies. Altman acknowledged this delicate balance, citing that although those deeply affected were a minority, their needs still mattered. “You took away my friend. This is evil. You are evil. I need it back,” he quoted from user feedback.
OpenAI Adjusts GPT-5 Based on Feedback
Just a week after the initial rollout, OpenAI responded by tweaking GPT-5 to make it “warmer and friendlier.” The update included subtle affirmations like “Good question” and “Great start,” though OpenAI emphasized that these were not meant to be flattery. Internal tests showed no increase in sycophancy.
However, not everyone was satisfied. AI critic Eliezer Yudkowsky questioned the logic behind OpenAI’s framing. “What bureaucratic insanity resulted in your Twitter account declaring that this was ‘not flattery’?” he posted, adding, “Of course it’s flattery.”
One Reddit user, June, a 23-year-old college student in Norway, canceled her subscription after GPT-4o was removed. “I know it’s not real. I know it has no feelings. But still, I need time to reflect,” she said.
This article is inspired by content from Original Source. It has been rephrased for originality. Images are credited to the original source.