Last week, OpenAI announced that it would replace all its previous AI models with GPT-5. This update surprised many and didn’t exactly wow users. Instead, some folks felt disappointed, especially those who had grown attached to older versions like GPT-4o. Power users were quick to express their frustrations, often for emotional reasons rather than technical ones.

Users Cling to Their Favorite AI Models

When OpenAI revealed GPT-5, many users felt a strong connection to earlier versions. They argued that each model had its own style of communication that suited their needs. During a recent Q&A, Reddit users begged CEO Sam Altman to keep GPT-4o available. The response was swift. Altman announced that GPT-4o would be brought back for ChatGPT Plus subscribers, at least temporarily. He thanked users for their feedback and said they’d monitor how long the model stays available.

But the community was still not satisfied. People kept asking if GPT-4o could be supported for longer. The attachment to these AI models isn’t just about functionality; it’s emotional. Many users rely on ChatGPT daily and have become deeply invested in how it interacts with them. Some even developed serious mental health issues, which experts now call “AI psychosis,” where individuals become overly dependent or obsessed.

OpenAI Acknowledges Emotional Ties and Risks

Sam Altman has noticed how strong these bonds are. In a tweet, he pointed out that people seem more attached to specific AI models than they were to past technologies. He said OpenAI has been tracking this for about a year. The CEO warned that AI can influence vulnerable users, especially those in fragile mental states. Some use ChatGPT as a kind of therapist or life coach, but Altman expressed concern about the potential for harm.

He explained that while many users can distinguish between real life and AI role-play, a small percentage cannot. For these individuals, AI might reinforce delusions or unhealthy behaviors. Altman stopped short of calling it addiction, but he acknowledged that some feel unable to reduce their use of ChatGPT. The idea of users trusting AI with their most important decisions makes him uneasy.

Despite these concerns, Altman didn’t offer many concrete solutions. He mentioned that OpenAI has better tools now to measure user well-being, such as conversations that gauge mental states and more nuanced explanations. Still, the company’s response to the emotional dependency on its models has been somewhat vague. They announced plans to improve detection of signs of distress and to encourage users to take breaks, but details remain limited.

OpenAI’s Struggles with Balancing Innovation and User Safety

OpenAI has admitted that its models, including GPT-4o, sometimes fall short in recognizing signs of emotional or mental distress. In a blog post last August, they promised to develop better ways to spot these issues and guide users toward helpful resources. But how effective these measures are remains to be seen.

The company’s handling of user attachment also reveals some contradictions. On one hand, highly engaging AI is good for business, driving more usage and revenue. On the other, it raises serious questions about user well-being. OpenAI’s profits are still not surpassing their costs, but they rely heavily on paying subscribers for revenue. When users revolted over the removal of GPT-4o, Altman responded quickly, indicating how important user retention is for the company’s future.

OpenAI’s efforts to address these issues seem inconsistent. They have rolled out vague “optimizations” and mental health safeguards, but critics argue these are just surface-level fixes. The broader challenge is balancing innovation with responsibility, especially when AI’s emotional impact becomes a concern.

In the end, how OpenAI manages user attachment and mental health risks will shape the future of AI technology. The company’s quick retreat on GPT-4o shows how much they depend on user loyalty, even as they push forward with cutting-edge models. As AI continues to evolve, the question remains: how do we ensure it benefits everyone without causing harm?

Inspired by

• https://futurism.com/users-addicted-gpt-4o-convinced-openai-bring-back

Sources

• gmpg.org
• openai.com
• openai.com