Chinese IT Companies Force Employees to Create AI Copies of Themselves

Chinese technology companies have crossed a line that was previously discussed only in theory: engineering department employees are directly asked to train AI agents that can perform their work. This is no longer an abstract threat of automation—it is a direct corporate mandate that places a worker in a fundamentally new position. In early April 2026, a project called Colleague Skill appeared on GitHub.

Its authors described the tool simply: it allows you to "distill" the skills and personality traits of colleagues to create their digital twin. In essence—to train AI to reproduce a person's professional behavior: their decision-making style, communication patterns, working heuristics. The project immediately attracted attention in the Chinese tech community and triggered a wave of public discussions that MIT Technology Review called a "painful self-analysis."

Notably, this wave spread not among technology skeptics, but among its main supporters. Chinese developers and engineers were generally among the most loyal to AI—they actively experimented with new tools, were first to implement them in workflows, and publicly promoted system capabilities. And now these same people found themselves in a situation where they were asked to become architects of their own replacement.

It is precisely this gap between beliefs and reality that caused such resonance. The situation exposes a fundamental contradiction in how companies present AI transformation. The narrative "AI is a tool that makes you more efficient" sharply conflicts with the corporate reality of the mandate "train an agent that will work instead of you."

The difference is not cosmetic—it changes the employee's very position. From a technology user, a person becomes its object: a data source from which value is extracted, rather than a partner in implementation. Similar signals are coming from other sectors of Asia.

Software development, content production, and technical support companies are increasingly testing schemes where a key employee task at a certain stage is to "upload" their knowledge to the system. After that, the agent takes over process management. This does not always mean immediate layoffs: some employers position such an approach as scaling capabilities.

But employees understand that their role in this model is finite. The Colleague Skill project made the invisible visible. Previously, similar processes were happening informally: companies analyzed work data, logged employee actions, studied decision-making patterns.

Now we are talking about an open, documented request: describe your decisions, systematize your heuristics, help build your digital copy. Such transparency is paradoxically more troubling than hidden data collection—it eliminates the possibility of maintaining illusions. An ethical question also arises, one that has barely been discussed publicly.

Does an employer have the right not only to demand the results of labor, but also to compel an employee to systematically transfer the very ability to produce them? In the labor law of most countries, there is no answer to this question. This is a fundamentally new form of alienation—not of the product of labor, but of cognitive capital itself.

For the labor market in the technology sector, this changes the basic equation of value. Traditionally, a specialist's tacit knowledge—accumulated experience, intuition, understanding of context and nuances—was considered difficult to reproduce. This is what made a qualified employee irreplaceable.

If this knowledge can now be systematically extracted and scaled through an agent, the protective barrier disappears. The response of the Chinese tech community showed: the boundary where enthusiasm about AI turns into existential anxiety turned out to be far thinner than expected. This is an important signal for the entire industry: the pace of automation has reached a point where abstract narratives about "reskilling" and "new roles" no longer work as reassurance.

The next question is not "will AI replace people," but "who is responsible for the replacement process."