Zhou ran quality assurance on AI models for a Chinese tech company. He spent his days catching the lies the language models told, signing off on the answers that were good, flagging the ones that hallucinated. Then his employer told him the AI was good enough to do his job, cut his salary from 25,000 yuan a month to 15,000, and showed him the door when he refused.

This week the Hangzhou Intermediate People's Court told the company to give him his job back. And the ruling is the first of its kind anywhere on the planet.

What the court actually said

Zhou was paid roughly 300,000 yuan a year — about $43,900 — to verify the accuracy of large language model outputs. When the company decided the LLM could verify itself, it didn't fire him outright. It demoted him with a 40% pay cut and called it "reassignment." Zhou said no. The company terminated him.

Under Chinese labor law, you can dismiss someone if there's a "major change in objective circumstances" that makes the contract impossible to keep. The company argued AI was that change. The court disagreed in unusually direct language. Adopting AI, the judges wrote, was a "voluntary move to stay competitive" — not an act of God. And by treating it like one, the company had "shifted the risks" of new technology onto its workers.

Zhou won at arbitration. The company appealed. He won at the lower court. The company appealed again. He won at the intermediate court too. The Hangzhou judges then published the case in a bundle of "typical examples" of AI labor disputes, timed for International Workers' Day on May 1. That timing is not a coincidence.

Why this matters outside China

The numbers in the rest of the world are ugly. In the first quarter of 2026, U.S. tech companies cut roughly 80,000 jobs. About 38,000 of those — 47.9% — were directly attributed to AI and workflow automation. Meta cut 8,000 staffers and Mark Zuckerberg said out loud that "compute and infrastructure" costs drove the layoffs. Oracle terminated up to 30,000 people in a single morning, including technical writers who had spent the previous year documenting their own workflows so the company's AI could learn from them.

None of those firings are illegal in the United States. There is no federal labor protection that says "you can't replace a person with a model." There is no state law that says it either. The U.S. Department of Labor has issued guidance. China's courts have issued judgments.

My Opinion

I'll be blunt. The country we keep calling authoritarian just gave its workers more legal protection from AI than any democracy on Earth. That should bother people who care about both labor rights and the rule of law, and I notice it doesn't seem to be bothering anyone.

Here's what bugs me. American tech CEOs will spend the next earnings cycle telling investors that AI productivity gains justify another 100,000 firings, and the same CEOs will spend the next political cycle telling regulators that any rule slowing them down is anti-innovation. Meanwhile a court in Hangzhou — not San Francisco, not Brussels, not London — looked at a guy who lost his job because his employer wanted cheaper output and said no, the cost of that decision belongs to the employer, not the human. That is, frankly, a more honest reading of who benefits from automation than anything the U.S. labor system has produced this decade.

The pattern from here is predictable. China will push automation as hard as anyone, but it will redirect the cost of displacement back onto firms through court precedent. The U.S. will keep socializing those costs onto laid-off workers and unemployment systems. And in two years some senator will hold a hearing asking why American labor markets feel hollowed out while Chinese ones don't, and someone will pretend the answer is mysterious.

Author: Yahor Kamarou (Mark) / www.humai.blog / 02 May 2026