The Cost of the Machine

How a Chinese Court Put Corporate AI Displacement on Trial

I. The Avatar

Somewhere in Shandong province, a game company built a ghost out of its former HR specialist. After the employee resigned, the company trained an AI system on every document she had written during her employment, fed it her work processing methods, gave it her photograph, and deployed it inside the company’s internal chat system. In the interface, the system introduces itself: “Hello, I am the digital avatar of a former employee. You can ask me questions at any time. I will respond based on the documents I worked with during my employment.”

The company called it a productivity experiment. The AI handles consultations, schedules interviews, builds presentations. It speaks in the employee’s patterns. It carries her institutional knowledge without her name on the payroll, without her salary, without her rights. The South China Morning Post reported the story in April 2026. It spread through Chinese social media within hours. The state broadcaster Xinhua cited it days later, in the same dispatch in which it announced the Hangzhou court ruling on AI-driven dismissal, as evidence of the broader anxiety the technology is producing inside a country that is simultaneously accelerating AI deployment and watching workers lose their positions to it.

The story of the Shandong avatar did not result in a court ruling. There was nothing to rule on: the employee had resigned and, according to reports, consented to the experiment. What it produced instead was a cultural rupture in a country where labor displacement is a political pressure point that the government tracks with the same attention it gives to currency reserves. The image that spread was not of a robot on a factory floor. It was of a desk job vacated, and the desk occupied by a facsimile of the person who used to sit there.

II. Zhou and the Legal Architecture

Three weeks after the Shandong story circulated, the Hangzhou Intermediate People’s Court published a set of case documents it had curated under the heading “typical examples of protecting the rights of AI enterprises and workers.” The publication date was April 28, 2026, three days before International Workers’ Day. The timing was institutional signaling.

The lead case involved a quality assurance supervisor identified by the court only as Zhou, hired in November 2022 at a fintech company in Hangzhou. His job was to evaluate the accuracy of outputs from large language models: reading what the AI generated, checking it against truth, flagging errors and privacy violations. His monthly salary was 25,000 yuan, roughly $3,640, or approximately $43,900 annually. He was not peripheral to the company’s AI operation. He was the human layer the company maintained on top of the machine because the machine could not yet fully supervise itself.

In 2025, the company decided the machine no longer needed that layer. It offered Zhou a reassignment: a lower-level position at 15,000 yuan per month, a 40 percent reduction. Zhou refused. The company dismissed him, calculating severance at approximately $45,000. Zhou contested the sum and the dismissal’s legality through labor arbitration. The panel ruled the dismissal unlawful and awarded higher compensation. The company filed a lawsuit at Yuhang District Court in August 2025. It lost. It appealed to the Hangzhou Intermediate People’s Court. In April 2026, it lost again. The court ordered it to pay Zhou over 260,000 yuan, approximately $38,067, in wrongful termination compensation.

The company had made a specific legal argument. Under Article 40 of China’s Labor Contract Law, an employer may terminate a contract when major changes in the objective circumstances on which it was based make it impossible to continue performing the agreement, provided the parties cannot reach a negotiated modification. The provision was constructed for situations where economic reality has been altered by forces beyond the employer’s control: company relocation, merger, regulatory collapse, force majeure. The company argued that AI-driven automation constituted exactly this: an objective major change in circumstances beyond its control.

The Hangzhou Intermediate People’s Court rejected the argument at every level of its construction. The court found that the company’s decision to implement AI technology was a voluntary business strategic choice, not an unpredictable or uncontrollable change in external circumstances. Beijing guidelines, issued by the Beijing Municipal Bureau of Human Resources and Social Security in December 2025, had already defined the threshold: objective major changes are events like disasters, forced closures, or regulatory ruination. Not investment decisions made in quarterly planning cycles. The Beijing bureau had released the definition as part of its annual compilation of typical arbitration cases for 2025, one of which involved a map data entry worker named Liu, hired in 2009, whose entire division was eliminated when his employer switched to AI-based data collection in early 2024. Liu won arbitration compensation on the same logic. The company’s AI pivot was a deliberate move to stay competitive. By citing AI as the grounds for dismissal, it had shifted the risks of its own technological iteration onto an employee who had no part in making the decision.

The Hangzhou court upheld that reasoning and added a second finding. The reassignment offered to Zhou was not a reasonable alternative. A 40 percent salary reduction, the court ruled, cannot be classified as a negotiated modification of the employment contract. It is constructive dismissal named differently. The court’s statement, published by Xinhua, put the operative principle in a single clause: companies cannot unilaterally lay off employees or cut salaries due to technological progress.

Wang Tianyu, a researcher at the Chinese Academy of Social Sciences, told Xinhua in the days following the ruling: “Technological progress may be irreversible, but it cannot exist outside a legal framework.” Pan Helin, an economist serving on an expert committee under China’s Ministry of Industry and Information Technology, argued that while AI-driven job displacement may be inevitable, companies must ensure fair treatment during transitions, including reasonable reassignment arrangements and adequate compensation. Wang Xuyang, a Zhejiang Xingjing law firm lawyer commenting to Xinhua, stated that while companies may benefit from AI-driven efficiency gains, they must also bear corresponding social responsibilities.

None of these voices were independent critics of the ruling. They were commenting within a framework the state was consolidating in public, three days before International Workers’ Day, through its own wire service.

III. The State’s Calculation

China’s 2026 government work report, the annual document in which the State Council sets national economic and social priorities, included for the first time an explicit directive on AI and employment: improving measures to promote employment and entrepreneurship in response to the development of AI. The inclusion is not symbolic. Government work reports are the formal mechanism through which national priorities are transmitted to provincial governments, regulatory bodies, and the judiciary. When a court in Hangzhou publishes a ruling on AI-driven dismissal three days before International Workers’ Day and three months after AI’s employment consequences entered the national policy document, the relationship between the ruling and the directive is not coincidental.

The state’s position is not opposition to AI adoption. China’s core AI industry exceeded 1.2 trillion yuan in 2025, with more than 6,200 registered enterprises. Official projections expect the penetration rate of next-generation intelligent terminals and AI agents to exceed 90 percent of the market by 2030. The state has staked technological sovereignty claims on AI leadership and has organized substantial industrial policy around those claims. What the 2026 government work report and the Hangzhou ruling do together is establish the terms on which AI adoption proceeds: companies may automate, but the transitional costs of automation are not available to be offloaded onto workers. The efficiency gains belong to the company. The liabilities of the transition belong to the company as well.

Behind this formulation is a problem that Beijing has been managing with increasing difficulty since 2023. Youth unemployment reached levels the government found politically uncomfortable enough to suspend publication of the monthly data. The broader economy has been sluggish under weight of post-pandemic contraction, property sector deterioration, and the secondary trade and energy disruptions produced by the Iran war. Corporate margins are compressed. In that environment, AI-driven displacement carries a specific political risk: it concentrates job losses among urban knowledge workers, the demographic with the highest expectations and the most visible presence in the social media environment through which public grievance circulates. A factory worker in an inland province whose job is automated by a machine is a labor market statistic. A fintech quality assurance supervisor in Hangzhou whose job is automated by a large language model is a case study, and when he wins in court, Xinhua covers it.

IV. What the Legal Clause Actually Forecloses

The consequences of the Hangzhou ruling extend beyond Zhou’s 260,000 yuan compensation. They close off a legal theory that Chinese employers had begun to use systematically.

The “objective major change” provision in the Labor Contract Law was designed as a narrow exception. Chinese labor law is weighted toward employment stability. Termination requires either documented worker misconduct, company-level economic distress, or a qualifying external disruption. The clause for objective major change, alongside provisions for mass layoffs tied to genuine financial difficulty, represents the employer’s strongest exit mechanism from a contract that has become operationally impossible. The company in Zhou’s case was attempting to expand this mechanism: it wanted AI adoption to qualify as an external disruption placing the employment contract beyond performance. Had the court accepted that argument, it would have opened a gap through which any company deploying AI could route its dismissals. Every termination associated with automation would have a legal classification waiting for it.

The court closed the gap by going to the origin of the clause. An objective major change is uncontrollable. It is unpredictable. It happens to a company, not inside a company’s strategy documents. The decision to purchase and deploy an AI system is the opposite: it is deliberate, planned, budgeted, approved by management, and executed on a timeline. Calling it an external disruption to the employment contract requires the company to argue that its own investment decision is something that happened to it. The court did not find this persuasive across arbitration, district court, and intermediate appellate review.

The Beijing bureau’s December 2025 guidance reinforced this by making explicit what the clause was never intended to cover. Subsequent legal commentary has noted that the ruling now places Chinese companies in a position analogous to where European employers operate under collective redundancy rules: AI adoption, if it results in workforce reduction, must be accompanied by a process. Retraining, genuine consultation, reasonable reassignment at comparable compensation, and severance that reflects the company’s gain from the technological transition rather than the statutory floor. The company cannot collect the upside and bill the worker for the displacement.

Greyhound Research’s chief analyst Sanchit Vir Gogia summarized the operative shift from a corporate governance perspective: the court narrowed an increasingly popular corporate shortcut, treating a voluntary automation choice as if it were an unforeseeable external shock that automatically justifies dismissal.

V. The Parallel Architecture of Displacement Elsewhere

In the first quarter of 2026, Challenger, Gray and Christmas recorded 217,362 job cuts announced across the American economy. Of these, 27,645 were explicitly attributed to artificial intelligence, constituting a full quarter of all layoffs tracked in March. The companies executing those cuts are not in distress. Meta projected capital expenditures of up to $145 billion this year. Amazon spent $44.2 billion on its cloud division in the first quarter alone. Block reduced its headcount from approximately 10,000 to fewer than 6,000; its CEO wrote in a shareholder letter that intelligence tools have changed what it means to build and run a company. Salesforce eliminated 4,000 customer support roles, with its CEO stating that he needed fewer heads. Oracle cut more than 10,000 positions, with savings allocated to data center funding. TrueUp’s layoff tracking puts tech sector job losses at 119,721 in 2026 through the end of April, a rate of 958 per day.

The stock market response has been documented and consistent. When companies announce AI-driven displacement alongside earnings, share prices rise. The financial incentive structure produced by this dynamic is not ambiguous: announcing workforce reduction framed as AI efficiency signals operational discipline to institutional investors. The worker displaced is credited with nothing. The executive who signed the termination notice may collect a performance bonus.

No US federal statute prohibits this. The Worker Adjustment and Retraining Notification Act requires 60 days’ notice before mass layoffs but contains no language prohibiting technology-driven displacement as a motive for termination. Employment-at-will doctrine in most US states removes the burden of justification from the employer entirely. The legal category the Hangzhou court deployed, the distinction between force majeure events and voluntary corporate strategic choices, has no equivalent in American employment law. A company in the United States that dismisses a worker because AI made the role cheaper to automate faces no legal challenge on those grounds. It faces nothing.

The EU AI Act, in force since August 2024, has produced the most elaborated regulatory framework in the Western world on AI in the workplace. Its core high-risk provisions for employers deploying AI in workforce management take effect in August 2026. Those provisions address transparency: worker notification, human oversight of AI-driven decisions, audit trails, bias monitoring. A 2025 University of Luxembourg analysis of the Act’s labor protections identified the gap explicitly. The framework asks whether AI was used responsibly. It does not ask who bears the cost of the decision to use it. A European employer who follows proper procedure can still dismiss a worker because automation made the role cheaper. The Hangzhou ruling says the procedure is not the point. The transfer of cost is.

VI. The Shandong Question

Return to Shandong. The gaming company’s HR avatar was not ruled unlawful. The employee had resigned voluntarily and consented to the data experiment. In the interface, the digital worker answers questions at any time, in the patterns of the person who trained it, paid for by none of the obligations that attached to her employment.

This is the condition the Hangzhou ruling is attempting to address before it becomes the norm: a company strategy in which the human worker is replaced not only by a general-purpose AI system but by a system trained on the specific worker’s own institutional knowledge, their methods, their language, their accumulated judgment. The replacement is not generic. It is personal. The company extracts everything the worker developed over years of employment, encodes it into a system it owns, and retains the output of that labor permanently without the labor costs.

The Hangzhou ruling speaks to the dismissal, not to what happens after. It protects Zhou’s right not to be fired on illegitimate grounds. It does not address the subsequent question, which the Shandong case makes visible: once the worker is gone, what rights govern the company’s use of what the worker produced? China does not yet have a settled answer.

Wang Tianyu’s statement to Xinhua, that technological progress cannot exist outside a legal framework, is doing more work than it appears to. It is not simply a gloss on the Hangzhou ruling. It is a description of the legal task that the ruling only begins. Courts can adjudicate individual dismissals. They cannot, without further legislative architecture, address the structural transfer of value from human labor to corporate-owned AI systems trained on that labor. The Hangzhou ruling establishes that Zhou cannot be dismissed for the company’s strategic choice. It does not establish what the company owes him for the knowledge it extracted and encoded before the dismissal.

That question has no courtroom yet.

VII. A Ruling, a Relay, and an Open Account

China does not operate a system of binding judicial precedent. Lower courts are not legally required to follow the Hangzhou decision. What the ruling creates is a signal, and in China’s legal architecture, signals published by a municipal intermediate court on the eve of International Workers’ Day, carried by Xinhua and state broadcaster CCTV, commented on by researchers from the Chinese Academy of Social Sciences and an expert committee of the Ministry of Industry and Information Technology, are not without operational weight. The December 2025 Beijing arbitration guidance and the April 2026 Hangzhou ruling are two points on a line. Whether additional courts extend it depends on how many Zhou-like cases arrive in arbitration panels over the next eighteen months, and whether the volume of disputes makes it politically untenable for local courts to rule in the other direction.

The government work report’s inclusion of AI employment effects as a national policy priority provides the political ceiling on that question. Beijing is watching. The courts know Beijing is watching.

What the ruling does not settle, because no ruling can settle it yet, is the technology’s trajectory. Judge Shi Guoqiang told CCTV that the court did not believe AI had reached the point of substantially replacing human workers. In 2026, across most job categories, he is correct. The Hangzhou ruling’s legal logic is calibrated to that reality: AI adoption is a strategic choice, not an objective transformation, because the transformation has not yet fully arrived. The provision for objective major change was designed for ruptures that leave contracts impossible to perform. When, in some categories of knowledge work, the rupture does arrive, when no reasonable reassignment exists because the entire function has been absorbed, the “impossible to perform” threshold will be litigated again, and the outcome will not be guaranteed by the 2026 ruling.

For now: Zhou filed, and won. Liu filed, and won. A gaming company in Shandong built a ghost out of its former HR specialist and deployed it in the chat interface. The court said the company that dismissed Zhou could not call its investment decision an act of God and send Zhou the invoice.

The deeper account, the one between corporate AI ownership and the workers whose labor trained it, remains open.

Sources

1. Hangzhou Intermediate People’s Court — Case documentation and published ruling — April 28, 2026

2. Xinhua News Agency — “China Focus: Chinese court defends labor rights in new AI-replacement case” — April 30, 2026

3. China.org.cn — “Chinese court defends labor rights in new AI-replacement case” — April 30, 2026

4. State Council Information Office (english.scio.gov.cn) — “Chinese court defends labor rights in new AI-replacement case” — April 30, 2026

5. CCTV — Interview with Judge Shi Guoqiang, Hangzhou Intermediate People’s Court — April 2026

6. Beijing Municipal Bureau of Human Resources and Social Security — Annual compilation of typical arbitration cases for 2025 — December 26, 2025

7. South China Morning Post — “Chinese firm slammed for using ex-employee’s data to create ‘AI human’ to continue working” — April 2026

8. ECNS / Chinanews.com — “Chinese firm tests AI ‘digital employees’ modeled on former staff” — April 7, 2026

9. Caixin Global — “Chinese Courts Rule Companies Cannot Fire Workers Simply to Replace Them With AI” — April 30, 2026

10. Wang Tianyu, Chinese Academy of Social Sciences — Comments via Xinhua — April 30, 2026

11. Pan Helin, Ministry of Industry and Information Technology expert committee — Comments via Xinhua / China.org.cn — April 30, 2026

12. Wang Xuyang, Zhejiang Xingjing Law Firm — Comments via Xinhua — April 30, 2026

13. Sanchit Vir Gogia, Greyhound Research — Commentary via CIO Magazine — May 2026

14. China 2026 Government Work Report — AI and employment provisions

15. Programs.com — “List of Companies Announcing AI-Driven Layoffs” — April 2026

16. TrueUp — Layoffs Tracker — accessed May 2026

17. Challenger, Gray and Christmas — Q1 2026 layoff data

18. University of Luxembourg — Law Research Paper No. 2025-01, “Labour Rights and the EU Artificial Intelligence Act” — January 15, 2025