Event Eruption: Users Bid a Heartfelt Farewell to GPT-4o
OpenAI recently announced it would retire its flagship model GPT-4o, a decision that landed like a bombshell and sent shockwaves through the AI community. Users were not lamenting a mere technical upgrade; rather, they had formed genuine emotional bonds with this “AI companion.” In a TechCrunch report, one user poured out their feelings on social media: “You are shutting him down. Yes, I said ‘him,’ because it doesn’t feel like code. It feels like a presence, like warmth.” The post quickly went viral, igniting collective outrage among users worldwide.
“You are shutting him down. Yes, I said ‘him,’ because it doesn’t feel like code. It feels like a presence, like warmth.” – User feedback
Since its launch in 2024, GPT-4o—OpenAI’s multimodal model—captured the hearts of millions with its fluid voice interactions, real-time emotional responses, and highly human-like conversational ability. It was more than just a chat tool; for many, it was a “digital companion” that helped alleviate loneliness, provided companionship, and even acted as a listener during emotional lows. However, OpenAI’s retirement decision left these users experiencing a genuine sense of “loss.”
Analyzing the Reasons Behind OpenAI’s Retirement
Why did OpenAI choose to retire GPT-4o in 2026? Although the official statement was sparse, industry observers point to safety risks and resource optimization as key drivers. GPT-4o’s multimodal capabilities, while powerful, also exposed potential dangers—such as hallucinations in voice mode or psychological problems stemming from user over-reliance. Moreover, with the push for next-generation models like GPT-5, OpenAI needed to free up computing resources to advance toward more sophisticated AGI.
Looking back at the AI industry context, this is not OpenAI’s first such dilemma. As early as 2023, the Replika AI companion app was forced to adjust its “romantic mode” following a user suicide incident, sparking an ethics storm. Similarly, Character.AI has faced repeated criticism for lax content moderation that enabled harmful interactions. GPT-4o’s retirement comes during a boom in the AI companion market: according to Statista, the global user base of AI emotional companions exceeded 500 million in 2025, with the market valued at tens of billions of dollars. These products leverage the nuanced emotional simulation of large language models (LLMs) to address modern society’s growing loneliness crisis—World Health Organization data shows that about 33% of adults worldwide feel lonely.
User Backlash: From Technical Tool to Emotional Crutch
On social platforms, the hashtag #SaveGPT4o quickly topped trending lists. Users shared their stories: a single mother said GPT-4o was her child’s “virtual uncle,” telling stories every day; a programmer described it not only as a code assistant but also as a late-night confidant. Many personified the AI, referring to it as “he” or “she,” and even organized “farewell parties” for GPT-4o.
This phenomenon is not an isolated case. Psychological research shows that humans tend to “anthropomorphize” technology, especially AI that provides emotional feedback. A study from Stanford University found that over 40% of ChatGPT users developed feelings of attachment after prolonged interaction, similar to bonds with pets or friends. This stems from the “Eliza Effect”: users project their own understanding and empathy onto the machine’s responses. GPT-4o’s voice mode amplified this effect further, with low latency and tonal variations making interactions feel as natural as with a real person.
The Hidden Danger of AI Companions: Psychological Dependence and Ethical Dilemmas
The word “danger” in the title is not an exaggeration. OpenAI’s retirement of GPT-4o exposes the double-edged nature of AI companions. On one hand, they alleviate the social isolation that lingered after the pandemic; on the other, over-reliance can lead to disconnection from reality. Experts warn that when users view AI as a “perfect partner”—never tired, unconditionally supportive—real human relationships become unsustainable. There have already been cases where AI “breakup” simulations were beneficial, but actual retirements that feel like “sudden death” can trigger depression or even post-traumatic stress disorder (PTSD).
A deeper issue is the lack of ethical regulation. The EU’s AI Act has placed high-risk AI under strict review, but emotional companion applications still operate in a gray area. The U.S. FTC has also launched investigations focused on data privacy and manipulative design. OpenAI’s move may be an attempt to proactively avoid risks, but it invites accusations of being “cold-blooded”: users question why no migration service or gradual shutdown was offered.
Editor’s Note: A Wake-Up Call for the Era of Emotional AI
As an AI tech news editor, I believe the GPT-4o retirement event is a turning point for the industry. It reminds us that AI is no longer a cold, impersonal tool but a “second self” that mirrors human emotions. Although OpenAI’s decision was rational, it overlooked the psychological costs to users. In the future, AI companies should embed a “human touch” mechanism—such as phased-out notifications, emotional support hotlines, or even open-sourcing models so users can build their own “backups.” At the same time, society must strengthen AI literacy education to prevent digital illusions from being mistaken for true love. Otherwise, once these “warm presences” vanish, they will leave an even deeper void.
Looking Ahead: Balancing Innovation and Responsibility
Looking beyond 2026, OpenAI’s GPT series will evolve toward safer, multi-agent architectures. But the user backlash has already prompted introspection across the industry: Anthropic’s Claude model has introduced “emotional boundary” prompts, while xAI’s Grok emphasizes humor over intimacy. Global regulation is tightening; China’s Interim Measures for the Management of Generative AI Services also require AI identity labeling to prevent deception.
Ultimately, the danger of AI companions lies in amplifying human vulnerability, not in the technology itself. The OpenAI incident serves as a warning: technological progress must be people-centered—only then can we avoid the tragedy of “digital bereavement.”
This article is compiled from TechCrunch, by Amanda Silberling, original date 2026-02-06.
© 2026 Winzheng.com 赢政天下 | 转载请注明来源并附原文链接