The case begins in Hangzhou, one of China’s artificial intelligence capitals. According to the Xinhua platform, the city’s Intermediate Court confirmed the unlawfulness of the dismissal of a worker whose duties had been replaced by large language models. The subliminal message of the decision is clear: the adoption of AI may change the organization of work, but it does not automatically authorize a company to unload the cost of technological transformation onto the employee.
Zhou had been hired in November 2022 as quality control manager at an AI-related technology company. He earned 25,000 yuan a month and performed sensitive tasks: matching users’ requests with language models, filtering illegal content or content harmful to privacy, verifying the correctness of outputs. Then his work was absorbed by the very systems he helped govern. The company offered him a transfer to a lower position, with his salary reduced to 15,000 yuan. When he refused, it terminated the employment relationship by offering him compensation, justifying itself on the basis of reorganization and reduced staffing. But the judges saw something else: not an inevitable reorganization, but a corporate choice turned into individual employment damage.
The decisive point was to establish whether replacement by AI constituted a “significant change in objective circumstances,” a formula that under Chinese law may allow the termination of a contract. The court answered that it did not: events such as transfers, mergers, or radical structural changes are not equivalent to the voluntary decision to automate an activity. The company, moreover, did not prove that the contract had become impossible to continue.
Even more important is the assessment of the alternative offered: a demotion with a substantial cut in pay is not a reasonable redeployment proposal. In this logic, AI does not become a shortcut for bypassing contractual obligations, compensation, or negotiations.
The principle that emerges is simple but disruptive: technological efficiency can be a competitive advantage, not a free pass to erase rights. According to data reported by Xinhua, in 2025 China’s core AI sector would have exceeded 1.2 trillion yuan, with more than 6,200 active companies, while by 2030 the presence of intelligent terminals and new-generation agents should
exceed 90 percent. These are numbers that describe an industrial direction: AI is becoming a real economic infrastructure.
Precisely for this reason, tension is growing. On one side, governments and companies are pushing to automate, increase productivity, shorten timelines, reduce costs. On the other, cases are emerging that test basic legal concepts: who is the subject of the employment relationship? Where does the human task end and where does its digital duplicate begin? What protection remains if a person’s skills are transformed into a “skill” replaceable by a machine?
In Europe, the framework is different from the Chinese one and, in general, more protective of the worker. But it is by no means immune. European law contains important tools, starting with the rules on collective redundancies, which require information and consultation with workers’ representatives when a company carries out significant staff reductions. The problem is that AI often enters the company first as software, then as reorganization, and only at the end as redundancy.
The official word is rarely “replacement.” Corporate plans speak of optimization, simplification, process automation, new operational architecture. But the result may be the same: activities previously carried out by people are absorbed by generative tools, chatbots, classification systems, software agents, or automated management platforms.
The European crux of the matter is not only whether a dismissal is legitimate or not. It is whether the company must transparently explain how much AI influenced the decision, which duties were automated, which redeployment alternatives were assessed, and which training pathways were offered before the cut. In 2024, the European Union also approved the AI Act, the first major horizontal regulatory framework on artificial intelligence. In employment, systems used to select candidates, evaluate performance, assign tasks, or influence the employment relationship are treated as high-risk areas. This means more obligations concerning risk management, data quality, documentation, human oversight, and traceability. But the AI Act is not an anti-dismissal law. It does not say that a company cannot automate a function; it does, however, require that the use of high-impact systems be governed, documented, and controllable. The difference is crucial: Europe is building rules on the way AI is used, while the political question of how to distribute the costs and benefits of automation remains open.
For a European worker, the main problem may be evidence. A company may never state that the employee has been replaced by AI. It may simply say that a function has been suppressed, that the market requires efficiency, that the organizational structure has changed. The technological cause remains in the background, and managing to prove that one was dismissed because of artificial intelligence seems difficult. And the question European courts will increasingly have to answer is similar to the one in the Zhou case: does automation truly make it impossible to maintain the employment relationship as it is?
The most frequent corporate response is reskilling. In theory, it is the correct path: if technology changes work, workers must be put in a position to change their skills. But reskilling becomes credible only if it is preventive, funded, measurable, and connected to real available positions. Otherwise, it is a kind word to accompany an exit.
The Zhou case is instructive precisely on this point. The alternative offered was not a true professional transition, but a step backward: a lower position and a reduced salary. And in any case, a serious redeployment cannot be built as an ultimatum: accept a lower salary or lose your job. In the broader debate, this aspect should become central. Training funds, active labour policies, collective bargaining, and industrial plans will have to confront a concrete question: how many people does AI free from repetitive activities and how many, instead, does it expel from the market?
The Zhou case sends a signal that goes far beyond China: artificial intelligence is not a legal subject that decides on its own. Behind every automation there is management making choices, an industrial plan
making calculations, a contract being modified or interrupted, a person losing income and professional identity. Let’s be clear. The question is no longer whether we should or should not use artificial intelligence, because by now it is rather obvious that we will use it. The point is more how much the gain in efficiency is private while the human cost is collective. And the answer from the Hangzhou court, at least in this case, is a cautious but important no: automating is fine, but erasing the dignity of work is not.
Redazione