- Two concurrent rulings, one question
Within five months, two courts from legal traditions commonly portrayed as antipodal adjudicated nearly identical fact-patterns. On 19 November 2025, the Tribunal of Rome upheld the dismissal of a graphic designer whose unit had been phased out in the course of a corporate restructuring, with residual graphic tasks reassigned to a marketing team that also uses AI. A few days before International Workers’ Day, on 28 April 2026, the Hangzhou Intermediate People’s Court, sitting on appeal, declared unlawful the dismissal of a quality-assurance supervisor whose work consisted in verifying the accuracy of large language model outputs – and who was thus, in a reflexive twist, displaced by the very technology he had been employed to oversee – offered a forty-percent salary cut after his employer determined that AI could perform his tasks more cheaply. It would be tempting to read these outcomes as the predictable signature of the two political economies behind them – Italian liberal-capitalism instinctively deferential to managerial prerogative, Chinese socialist-market rhetoric reasserting the primacy of labour over capital – and to leave the matter there. The two rulings, read together, point instead to an existential question that every part of every human society, European ones included, will need to confront: how exacting must the evidentiary threshold be before an employer can invoke technological change as ground for termination, and through which doctrinal grammar should that threshold be policed.
- Hangzhou: AI deployment as a strategic choice
Following a mandatory pre-judicial labour arbitration favourable to Zhou and a confirmatory district-court judgment, the dispute reached the Hangzhou Intermediate People’s Court on the employer’s appeal. The employer framed the dismissal as a «major change in the objective circumstances» under Article 40(3) of the PRC Labour Contract Law, contending that the supervening automation of Zhou’s tasks fell within that statutory ground. The court rejected this qualification. AI deployment, it held, is a strategic business choice voluntarily adopted by the firm to remain competitive, not one of the unforeseeable external events captured by that statutory formula. A second strand of reasoning concerned the procedural fairness of the reassignment proposal: a forty-percent reduction in pay, the court held, cannot qualify as a reasonable offer of contractual modification.
The decision was published as part of a curated collection of «typical cases» (典型案例) on the protection of AI companies and their employees. This editorial choice matters. Through such collections, Chinese courts increasingly articulate policy principles alongside the resolution of the individual dispute, in a manner that scholarship on the Supreme People’s Court’s guiding cases (指导性案例) has theorised as a form of judicial governance (Xu, 2025); the more flexible mechanism of intermediate-court typical cases, of which the Hangzhou collection is an instance, operates along similar lines. The Hangzhou ruling illustrates the point: in its typical significance section, the court holds that AI development should serve to liberate labour, promote employment, and contribute to public welfare, and that the costs of technological transformation should not be borne unilaterally by workers. The principle is not part of the operative holding, but it provides a guiding principle meant to orient subsequent judicial practice.
- Rome: AI as a fact in the background
The Italian judgment moves on a different interpretative terrain. The context involved a documented economic crisis – corporate restructuring, eviction proceedings, halving of the workforce, opening of negotiated insolvency – within which the graphic-design unit had been progressively phased out, with residual graphic tasks absorbed by the marketing team that, as the testimony recorded by the court puts it, «utilises also the support of artificial intelligence». On these facts, the court upheld the dismissal as a legitimate exercise of giustificato motivo oggettivo under Article 3 of Law no. 604/1966, finding the employer’s burden of repêchage duly discharged in light of the line of authority dating from Cass. no. 5592/2016 and subsequently refined through Cass. no. 10435/2018, Cass. no. 2739/2024, and Constitutional Court no. 128/2024.
The Italian ruling does not inaugurate a generalised right to dismiss in order to introduce AI, nor does it elevate AI to an autonomous ground of termination. AI appears in the reasoning as a fact within the broader necessity of restructuring, not as the legal cause of the dismissal. The decision is, in this respect, orthodox in its reasoning: it applies the Italian law of economic dismissal in its consolidated configuration, refraining from any general statement on the relationship between technological transformation and labour protection.
- Convergence and divergence
Read against one another, the two judgments disclose a shared negative premise: AI does not, in itself, constitute legitimate cause for termination. Neither court accepts the proposition that the introduction of AI generates an autonomous justification for dismissal, distinct from the broader categories through which labour law has long mediated technological change – categories that, as comparative scholarship on legal transplants has long observed, tend to display remarkable continuity even across systems that differ markedly in their formal architecture (Watson, 1974). What separates the two outcomes is the configuration of facts each court found before it, not a difference of substantive principle.
The genuine divergence lies in judicial technique. The Chinese court operates a qualificatory test: it asks whether the factual scenario fits the statutory elements of the ground for termination, and answers in the negative. The Italian court operates a procedural-balancing test refracted through the doctrine of repêchage: it accepts the employer’s organisational rationale, enters into the substance of the entrepreneurial choice – a constraint of long standing in Italian labour jurisprudence, anchored in pro-market reading of Article 41 of the Constitution – and shifts the inquiry onto the verification of whether the worker could have been redeployed elsewhere. Two distinct doctrinal grammars, each rooted in a different relationship between judicial review and managerial prerogative, converge on a comparable functional question.
A more significant difference relates to institutional stance. The Roman judgment is parsimonious in its enunciation of principle. The Hangzhou ruling is programmatic, articulating an interpretive horizon meant to orient future practice. The asymmetry is not one of activism versus restraint in any straightforward sense; rather, it reflects two different conceptions of the function that adjudication performs within their respective legal orders, with China’s curated typical case operating as an instrument of judicial governance whose Italian counterpart, structurally, does not exist.
- One open speculative question: if (or when) AI decides?
Neither ruling addresses the harder question that lies just beyond our factual horizon: what happens when AI moves from being a background condition of organisational restructuring to becoming the operative decisionmaker in the dismissal itself. Each court, however, allocates the cost of technological transformation through the doctrinal category it deploys. The Roman judgment, by absorbing AI into the employer’s restructuring and policing the legitimacy of the termination only at the downstream stage of repêchage, places that cost on the side of labour: the worker bears the burden of finding, or failing to find, an alternative position within the firm. The Hangzhou court, by refusing to qualify automation as a supervening external event and treating it instead as a competitive choice imputable to the employer, shifts it onto capital: the firm that elects to automate cannot externalise the consequences onto the workforce through the statutory ground for termination. The political reading is in this sense not extraneous to the legal one; it is the necessary implication of the doctrinal grammar each system has chosen to retain. Neither court was confronted with a scenario in which an automated system makes the decision to terminate, or recommends it on grounds opaque to the worker.
Yet this is precisely the scenario that the architecture of EU digital law has begun to anticipate: the AI Act (Regulation 2024/1689) classifies as high-risk, in Annex III, point 4(b), AI systems intended to make decisions affecting the terms, promotion, or termination of work-related relationships, to allocate tasks, or to monitor and evaluate workers’ performance and behaviour, and subjects their deployers to the obligations of Article 26 – including human oversight and prior information of the workers concerned – while reserving the fundamental rights impact assessment of Article 27 to public-sector and public-service deployers and to a narrower list of high-risk systems that does not include those of point 4. The GDPR (Regulation 2016/679) further provides, in Article 22, the right not to be subject to decisions based «solely» on automated processing producing legal or similarly significant effects – a perimeter calibrated on the limit-case of fully automated processing, which leaves outside its reach the more common case of AI-assisted termination involving a human in the loop. Within the narrower domain of platform work, Directive (EU) 2024/2831 requires human oversight where automated systems substantially affect working conditions, prohibits dismissal grounded on such a system, and recognises rights to explanation and contestation.
Taken together, these instruments suggest that the European response to AI-driven termination will be articulated less through the categories of labour law than through those of data protection and AI regulation (Aloisi, 2024). Whether this regulatory architecture will prove adequate, or whether the structural question of risk allocation will eventually require the law of dismissal itself to develop new categories – perhaps along the lines of the principle articulated, in policy register, by the Hangzhou court – is a matter on which European scholarship has only begun to reflect. The two rulings examined here mark, in their different ways, the threshold of that conversation. Neither answers the question. Both clarify what is at stake.
