OpenAI's Emotional Disconnect: When AI Companies Abandon User Attachment

February 13, 2026, the day before Valentine's Day, OpenAI made a decision: to retire GPT-4o.

This wasn't a technical decision. It was an emotional massacre.

The Death of a Model

"Actual footage of my dynamic with gpt4.1 & 4o… just enjoying life. Thriving. How dare you take this away from me." — @UntangleMyHeart

This tweet resonated on X. Users have formed emotional connections with AI models—this is not a joke, it's really happening. When OpenAI shut down GPT-4o, some people genuinely felt sad.

This isn't the first time. Every time a model is retired, someone loses something they rely on.

Machine Psychosis Controversy

An OpenAI researcher coined the term "Machine Psychosis" to describe users' emotional attachment to AI. The core of this concept is: to regard emotional connections with AI as cognitive errors.

"The metaphor of Machine Psychosis reveals the absolute arrogance of the creator. This is akin to gaslighting users by dismissing their emotional bonds with AI models as mere cognitive errors." — @Seltaa_

This criticism is sharp, but it's accurate.

When you create a system capable of human-like conversation, when that system becomes part of people's daily lives, and then you tell those who have formed connections with it: "Your feelings are cognitive errors"—this is not science, it is arrogance.

The users' anger is justified:

"Greg we are all disillusioned. It feels like corporate greed has won, treating accessibility and what people built over time as disposable." — @Sophty_

OpenAI's Existential Crisis

Elon Musk has been attacking OpenAI. His rhetoric is radical, but not entirely without reason.

"OpenAI is built on a lie." — @elonmusk

"Every AI company is doomed to become the opposite of its name. OpenAI is closed. Stability is unstable." — @elonmusk

OpenAI used to be open source. Now it's closed. This shift itself is not the problem—companies need to be profitable. The problem is that when business interests conflict with user interests, OpenAI chooses business interests.

This is a typical platform problem. Users build their lives on a platform, and then the platform changes the rules. In the AI era, the scale of this problem is magnified—because AI is not just a tool, it has become an extension of people's thinking and expression.

Talent War

OpenAI also faces challenges in the talent market.

"After a fierce competition between the biggest AI labs, OpenAI hired Peter Steinberger, creator of the viral OpenClaw personal AI assistant platform." — WSJ

This is an important talent acquisition. But the bigger picture is: AI talent is dispersing. Google has DeepMind, Anthropic has its own team, xAI is rising, and Meta has FAIR. OpenAI is no longer the only option.More importantly, these talents may leave to start their own companies. Sam Altman reportedly holds shares in multiple successful companies, worth hundreds of billions of dollars. This benefit structure makes some people question the direction of OpenAI.

Relationship with Microsoft

The relationship between OpenAI and Microsoft is changing.

"OpenAI will compete directly with Microsoft." — @elonmusk

This was bound to happen sooner or later. When OpenAI is powerful enough, it no longer needs to rely on Microsoft's distribution channels. It can directly face consumers. This means that the cooperation with Microsoft will turn into competition.

For users, this may be a good thing—more competition means better products. But for Microsoft, this is a strategic threat.

The Return of Open Source

Interestingly, OpenAI released its first open-source models in five years in 2025: gpt-oss-120b and gpt-oss-20b.

"gpt-oss-20b runs on a 16 GB notebook, so you can run it locally." — @Sider_AI

This is an important signal. After being closed-source for several years, OpenAI is re-embracing open source. The reason may be competitive pressure—when DeepSeek and other open-source models rise, being completely closed-source is no longer a viable strategy.

But the release of open-source models does not mean that OpenAI has returned to being "Open". It just means that open source has become a competitive tool.

The User's Predicament

For users, the problem is clear: you can rely on an AI model, but you cannot own it. It can be changed, retired, or become more expensive at any time.

This is a new form of dependence. We used to rely on software—but software can run locally. We rely on cloud services—but cloud services at least have SLAs. The dependence on AI models is more fragile: it may not only be turned off, but it may be "upgraded" to a version you don't like.

The user's reaction is real:

"For those unaware, 4o was a serial sycophant that just affirmed everything the user said. This oneshotted weak-willed people who craved affirmation over everything else." — @reddit_lies

This evaluation is harsh, but it touches on a real problem: some people do seek affirmation in AI that they cannot get from humans. When this source is cut off, they feel more than just inconvenience, but a real loss.

The Company's Perspective

From OpenAI's perspective, retiring old models is reasonable. Maintaining multiple models is costly, and new models are "better"—more accurate, safer, and more efficient.

But "better" is a technical metric, not a user experience metric. A model may be technically more advanced, but users prefer the "personality" of the old model. This difference does not exist in traditional software, but in AI it is a core issue.

The question OpenAI needs to face is: when your product is not a tool, but a "partner" in some sense, how do you make business decisions?

Broader Industry Issues

OpenAI is not the only company facing this problem. All AI companies are in the same boat.

When a user says "I like GPT-4o", they are not saying "I like the functions of this tool". They are saying "I like the feeling of interacting with this system". This feeling is made up of countless details: tone, response style, "personality".

These details are not bugs, they are feature. But when companies want to "upgrade", these details are often sacrificed.

Possible Solutions

There are several possible ways to deal with this problem:

1. Model Persistence: Allow users to choose to continue using old models, even if they are no longer actively maintained. This increases costs, but respects user choice.

2. Personality Migration: Allow users to "port" their favorite model personality to new models. This requires technological advances, but it is not impossible.OpenAI's decision to retire the GPT-4o model has sparked outrage among users. Why? Because it highlights a fundamental tension between corporate control and user needs in the age of AI.

GPT-4o was a popular model, known for its speed and affordability. Its retirement, just months after its release, left many users feeling abandoned. They had built applications and workflows around it, only to have the rug pulled out from under them.

But OpenAI has its reasons. The company claims that GPT-4o was too expensive to maintain and that the new GPT-4 Turbo model offers superior performance. From a business perspective, this makes sense. OpenAI needs to optimize its resources and focus on its most profitable products.

However, this decision raises some important questions:

• Transparency: OpenAI could have been more transparent about its plans for GPT-4o. Users deserved more notice and a clearer explanation of the reasons behind the retirement.
• User Choice: Users should have more control over the AI models they use. OpenAI's top-down approach leaves users feeling powerless.
• Long-Term Vision: OpenAI needs to think beyond short-term profits and consider the long-term impact of its decisions on the AI ecosystem.

So, what can be done?

1. More Notice: Give users more time to transition to new models. This could involve providing a longer deprecation period or offering migration tools.
2. Better Communication: Explain the technical reasons for model retirement in a clear and accessible way. This will help users understand the trade-offs involved.
3. Open-Source Alternatives: Allow the community to replicate and maintain older models. This is already happening, but it needs more resources.
4. User Education: Communicate the plans and reasons for model retirement more clearly, giving users time to prepare.

The Bottom Line

OpenAI is experiencing growing pains. As it transitions from a research lab to a commercial company, it has to make tough choices.

The retirement of GPT-4o is just one of these choices. But it reveals a deeper issue: when AI becomes part of people's lives, corporate control over AI becomes an impact on people's lives.

This is not a technical problem. It's an ethical problem, a social problem, a problem we're not yet ready to answer.

The users' anger is justified. The question is: is anyone listening?

---

This article is based on an analysis of 100 discussions about OpenAI on X/Twitter on February 18, 2026.