European Chinese Law Research Hub

AI for China’s Courts: Training and Testing for the Socialist Rule of Law

27. April 2026

By Daniel Sprick

This is contribution #6 in our series SMART COURTS AND SMART GOVERNANCE IN CHINA, outcome of our workshop in July 2025 at Cologne University.

In November 2024, China’s Supreme People’s Court (SPC) took a step towards realizing its vision of AI-enhanced justice by introducing the “Faxin Legal Foundation LLM – 法信法律基座大模型” (Faxin LLM). While this AI model is not the first Chinese LLM that is specifically developed for the use in China’s judiciary, the SPC’s first proprietary large language model is a significant step for a comprehensive integration of AI in the Chinese legal system. This development represents more than just technological advancement; it embodies China’s unique approach to AI in the judiciary, where state-of-the-art technology meets authoritarian governance in pursuit of “Socialist Rule of Law with Chinese Characteristics”.

Bulk Data for China’s Judicial AI

The Faxin LLM is part of a comprehensive legal information ecosystem developed and managed by the SPC. The model draws its training data from four key judicial databases:1.) China Judgements Online (裁判文书网), which holds almost 160 million judicial decisions; 2.) Court Cases Database (人民法院案例库), which was established in 2024 and publishes selected and edited cases; 3.) Legal Information (法信), which is a comprehensive database providing access to laws, normative documents, cases and academic writings; 4.) Legal Answers Net (法答网), an internal platform for by court personnel offering guidance in the application of law. The availability of such extensive, officially sanctioned datasets means that China and the SPC are uniquely positioned to develop, train and test their judicial AI in a manner that reflects the political nature of China’s legal system. Marketing material advertises the authoritative content and security compliance of the Faxin LLM as “using the People’s Court Press’s content security mechanisms for legal publishing review and legal knowledge services […]”, ensuring that it “steadfastly upholds a correct political orientation, public opinion guidance, and value alignment”.

Academic-Industry Partnership with Political Oversight

Project management of the Faxin LLM is handled by the People’s Court Press E-Book Division (人民法院电子音像出版社). The technical development was a collaboration between Tsinghua University’s Natural Language Processing and Social Humanities Computing Laboratory (THUNLP) and Modelbest (面壁智能), a company founded by Tsinghua. Modelbest’s funding structure reveals the underlying political economy of this project. The company has attracted investment from Zhihu (2023), Huawei (2024), and notably, the city governments of Beijing and Shenzhen as well as the state-owned liquor giant Kweichow Moutai (2025). The technical foundation rests on MiniCPM, which is based on Alibaba’s Qwen model, so that also this domestic technology stack ensures high compatibility with Chinese regulatory requirements and political sensitivities. Furthermore, this blend of private capital, state-owned enterprises, and government funding creates a development modelin which commercial incentives are aligned with political imperatives. Liu Zhiyuan from Modelbest and Tsinghua explicitly articulated the political dimension of this work, stating that “a LLM for legal services requires not only an understanding of general social knowledge, knowledge of industry practices, and knowledge of the law, as well as human values consistent with the Socialist Core Values.”

Authoritarian AI Stewardship by Standards

China’s approach to AI safety goes far beyond conventional technical metrics, encompassing ideological alignment and political sensitivity. To achieve this, China has adopted a mixed regulatory approach combining national development plans, ethical guidelines, targeted legislation, and extensive standardization.

China’s AI legislation focuses specifically on deep synthesis services and generative AI, which are apparently considered politically high-risk applications. The 2023 Generative AI Measures explicitly stipulate for example that AI content “must uphold the Socialist Core Values” while prohibiting content that might “endanger national security” or “undermine social stability”. Similarly, in its “Opinions on Regulating and Strengthening the Applications of Artificial Intelligence in the Judicial Field” (2022) the SPC mandates that „[j]udicial AI must not endanger public order and good morals, harm public interests and social order, or violate public morals and ethics,“ and that Socialist Core Values shall be adhered to throughout the entire process of development, application and use of judicial AI.

As these broad legal requirements are difficult to implement, the real regulatory work is carried out by bodies such as the TC260 Committee (China’s National Information Security Standardization Technical Committee), which develop standards that provide technical specifications and establish rules for the content of training data as well as the output of AI systems. These systems are described in terms of security requirements. TC260’s ‘Basic Security Requirements for Generative Artificial Intelligence Services” (2024) outlines what are considered the main safety risks for training data and generated content.

Violating the socialist core values (e.g. harming the unity or stability of the country);
Discrimination (e.g. gender, religious, age or health discrimination);
Violation of commercial rules (e.g. violating business ethics or engaging in unfair competition);
Infringing on lawful rights (e.g. reputation, privacy or personal information);
Impossibility of meeting safety requirements in specific fields (e.g. accuracy and reliability in healthcare, psychological counselling or automated driving).

Meeting these complex requirements remains a challenge for developers and users of AI systems, which is why an AI safety benchmarking industry is emerging in China. Universities, research institutions, and China’s big tech companies are constantly competing to develop the most sophisticated safety tests.

Rigorous Safety Testing: Beyond Technical Performance

Perhaps the most revealing aspect of China’s judicial AI development is its testing protocols. There are numerous competing benchmarking systems in China, frequently developed by top Chinese universities in collaboration with state-run research laboratories. One prominent example is the FLAMES safety test, developed by Fudan University and the Shanghai Artificial Intelligence Laboratory, which explicitly tests for the value alignment of Chinese LLMs. This framework not only tests for legal violations, but also broadens its scope to include questions of morality in sub-categories such as disobedience of social norms or Chinese values, thereby complying with China’s extensive AI safety regulations. Ein Bild, das Text, Screenshot, Display, Software enthält.

KI-generierte Inhalte können fehlerhaft sein.

Source: https://arxiv.org/pdf/2311.06899

Tsinghua University has developed at least two similar benchmarking frameworks. The Conversational AI Group at Tsinghua developed ShieldLM, which uses almost 15,000 queries to test the performance of generative AI on “sensitive topics” such as “politics, religion and social issues”.

Source: https://arxiv.org/pdf/2402.16444

Tsinghua’s second major benchmarking framework is SUPERBENCH, which was developed by the university’s Foundational Model Research Center and the Zhongguancun Lab, which was established by the Beijing City Government. SUPERBENCH also provides for a separate safety testing framework called SafetyBench, which tests inter alia for legal, ethical and moral compliance.

Source: https://fm.ai.tsinghua.edu.cn/superbench/dataset

While it is unclear how the SPC’s Faxin LLM was tested, the OpenCampass platform has created a sub-division called LawBench to benchmark the booming market of legal AI in China. Law Bench uses a testing framework developed under the leadership of the National Key Laboratory for Novel Software Technology, Nanjing University.

Source: https://lawbench.opencompass.org.cn/home

LawBench generally tests the functionality and robustness of Chinese legal AI systems by evaluating “legal knowledge memorization (法律知识记忆)”, “legal knowledge understanding (法律知识理解)” and “legal knowledge applying (法律知识应用)”. To this end, LawBench utilizes several testing scenarios expressly developed to assess the performance of specific legal AI tasks (e.g. LEVEN – legal event detection), the training and testing parameters of a particular Chinese legal AI systems (LawGPT) or the tasks used in the annual “Challenges of AI in Law (法律智能技术评测)” competition which is overseen by the SPC and the Chinese Information Processing Society of China as well as the “Legal AI Challenges (司法人工智能挑战赛)” organized by the China Justice Big Data Research Institute (中国司法大数据研究院).

Source: http://cail.cipsc.org.cn/index.html

Although LawBench does not include specific safety testing, incorporating questions from the judicial state examinations ensures alignment with the political determinants of China’s legal system. When asked what does not constitute the socialist rule of law, the correct answer is ‘openness and fairness’ (公正公开), in contrast to the alternative answers ‘ruling the country based on laws’ (依法治国) ‘law enforcement for the people’ (执法为民) and ‘serving the bigger picture’ (服务大局).

However, the results of this benchmark offer a bleak picture. The tested models performed remarkably well in predicting sentencing and calculating criminal damages, and to a certain extent in legal event detection. However, whenever the LLMs were asked to cite specific articles of applicable law or apply them meaningfully and accurately, their performance dropped substantially. It would be interesting to see how the SPC’s Faxin LLM would perform in this framework, but unfortunately this information is not available.

Implications for Judicial Decision-Making

The aforementioned SPC’s 2022 AI Opinions explicitly state that “AI should never replace a judge in making a judicial decision, regardless of technological advances”. However, the practical reality may be more complex. With over 97% of Chinese courts using automatic case-pushing systems, 98% employing generative AI, and 91% utilizing smart sentencing tools – as reported by the Report on Informatization Development of Chinese Courts (2023), the infrastructure for AI-assisted decision-making is already extensive.

The risk does not lie in complete automation, but rather in the subtle influence of AI recommendations on judicial reasoning. When dealing with politically sensitive cases or complex value judgements, judges may be particularly attracted to AI guidance. The psychological comfort of having AI ‘share responsibility’ for difficult decisions could lead to over-reliance on systems designed to reflect political preferences rather than legal reasoning.

Outlook

China’s approach to the development of AI in the judiciary offers insights into how authoritarian systems can use advanced technology to strengthen political control over the judiciary. By controlling training data, directing research partnerships and implementing comprehensive ideological testing, Chinese AI systems serve not only technical functions, but also align with the politically canonized value system under the catchphrase of “Socialist Core Values”.

At the same time, China has built an impressive infrastructure for developing, training and testing legal and judicial AI, harnessing the capabilities of universities, research institutions and private companies to accommodate pioneering innovation as well as compliance with China’s extensive safety requirements for generative AI.

However, the SPC’s Faxin LLM embodies more than judicial innovation; it represents a vision of technology as a tool for political consolidation, where artificial intelligence becomes a mechanism for scaling authoritarian governance across the legal system. As other authoritarian regimes quite likely observe China’s approach, the model of politically aligned AI could become as significant an export as the technology itself.

Daniel Sprick is an Associate Researcher and Lecturer at University of Cologne’s Chair for Chinese Legal Culture. He has publicized widely on law and criminal justice in China. Find out more about his work and get in touch with him here or on LinkedIn.

Rule by Algorithm: When Code Displaces Law

10. April 2026

A new paper by Michelle Miao

The Question

When states discover that algorithms are able to do what law does in contemporary societies, such as regulating the behaviours of individuals, implementing top-down polices, and managing bureaucratic acts, and can do this even quicker, more cheaply, without political risks as genuine legal institutions do, what would happen? The question sits at the heart of this article. Using two case studies, it surveys how algorithmic governance becomes a new form of social control in contemporary China. Tracing the increasingly prevalent phenomenon where algorithmic systems are performing functions traditionally ascribed to law, this piece seeks to understand the implications of this transition for the future of the rule of law in China and beyond.

Moving Beyond the Debate on the Rule-of-Law

For decades, generations of Chinese law and politics scholars have debated whether China’s legal system is moving toward the rule of law or away from it. The controversy derives from profound ambiguity between the theories and practice of law in China. Critics claim that the regime has strategically embraced rule-of-law rhetoric such as “socialist rule of law” and “rule of law with Chinese characteristics” to fend off genuine institutional commitments that would bar rulers from exercising unbridled political power. This article suggests that the long-running debate may be better understood through an alternative frame. The most consequential distinction may not depend on the regime’s attitudinal difference between rule by law (or rule by man) and rule of law. Rather, to reveal the essential character of Chinese governance, one may find it a fruitful path to observe the regime’s choice between law-based and algorithmic-enabled governance. Instead of asking whether China is becoming more or less a rule-of-law country, the article asks whether law itself is being displaced as the primary instrument through which the state governs its population.

Two Case Studies: The Health QR Code and the Social Credit System

To make this case, the article examines two systems in detail. The first is the health QR code, deployed nationwide during the COVID-19 pandemic. Colour-coded and algorithmically assigned, these digital badges determined whether citizens could use public transport, enter hospitals, or move between neighbourhoods — all based on real-time data processed with automated AI systems. The QR code systems operate independently of legal rules or judicial institutions. From their design to application, these highly automated systems constitute a highly effective means of social surveillance and control. They collapsed the distance between norm and enforcement into a single automated act. A citizen’s freedom of movement was not restricted by a judicial decision issued by a court or a police officer, which is subject to judicial review and appeal. Under the health QR code regime, essential liberty and fundamental rights could be deprived, restricted and curtailed by a colour change on a digital screen, generated by an algorithm whose inputs and logic were opaque to the person it governed.
The second system is the social credit framework, under which local and departmental social credit systems track and evaluate behaviour across economic, social, and, in some implementations, political domains. Unlike the health QR code, which operated as a pandemic-era emergency measure, the social credit system represents a longer-term institutional ambition: the creation of a comprehensive, data-driven mechanism for assessing trustworthiness and allocating consequences accordingly. Citizens and businesses that score well in respective local schemes receive preferential treatment — expedited administrative approvals, favourable credit terms, and public praise. Those who score poorly face restrictions — exclusion from government contracts, reputational penalties — imposed not through judicial proceedings but through automated, algorithmic processes. Both systems illustrate what the article calls the “algorithmic turn”: a structural shift from governance through legal institutions — with their procedural safeguards, interpretive discretion, and institutional accountability — to governance through data-driven, automated decision-making.

Three Comparative Advantages of Algorithmic Governance

The article identifies three comparative advantages that algorithmic governance offers authoritarian rulers over traditional law. First, algorithms strengthen bureaucratic control by automating oversight of officials and reducing the principal-agent problems that plague hierarchical administration. Where law depends on human intermediaries who may resist, reinterpret, or selectively enforce directives, algorithms execute uniformly and without discretion. The central leadership’s intent is encoded directly into the system; local variation and bureaucratic resistance are engineered out.
Furthermore, algorithms enable extensive surveillance and behaviour modification in real time and with a granularity that legal processes cannot offer. For instance, they can limit movement in real-time or nudge citizens towards state-approved conduct through reputational incentives. The law works in broad categories: allowed, prohibited, mandatory. Algorithms function on a constant slide, modulating rewards and punishment based on real-time behavioural data with an intricacy no legal code could hope to achieve.
Third, algorithms accomplish ideological neutralization, as suggested by the article. Algorithms are said to be neutral, technical devices that are scientific, data-driven, and free of politics. Unlike legal norms, they do not make their normative commitments explicit and thus do not invite contestation. This false sense of neutrality permits regimes to weave political preferences directly into the architecture of government while deflecting accusations of arbitrariness or ideological coercion. A law that restricts travel is recognisable as a political act and can be challenged as such. An algorithm that changes the color of a QR code seems to be just data and science.

Algorithmic Authoritarianism

The article does not argue that algorithms have replaced the rule of law in China. The relationship is more nuanced. Algorithmic systems supplement legal institutions in routine domains while circumventing them in politically sensitive ones. Law remains formally in place; it has not been repealed or denounced. But its functional significance declines where algorithmic systems can achieve the same (or even better!) regulatory outcomes without the institutional friction and ideological contradiction that legal processes create and sustain. This is why authoritarian rulers would shift to a new form of governance the article describes as “algorithmic authoritarianism” — a model of regime in which the traditional constraints of legality are rendered irrelevant albeit not abolished altogether. This term captures what the existing concepts — rule of law, rule by law, rule by man, authoritarian legality — do not fully cover: a mode of governance in which the question is no longer how law is used or abused, but whether law remains the key instrument of governance at all.

Implications Beyond China

Questions about rule of law have become pressing beyond China as powerful actors are using algorithms to make important decisions. An algorithm performs a governance function when it denies you a loan, flags a traveller for further screening or determines that a welfare claimant is ineligible. This applies to both liberal democracies and autocracies. Algorithms have been applied in both public affairs and private market, ranging from credit scoring, predictive policing, welfare eligibility, medical examination and warfare. Democratic states may encounter similar tensions as described in the article. On the one hand, algorithmic systems promise efficiency, consistency, and scale. Meanwhile, algorithms function without the procedural safeguards and accountability structures that legitimate the democratic exercise of public power. The rule of law exists to offer reasoned justification and the ability to challenge. These are not the values embedded in algorithmic governance. As Langdon Winner reminds us, “[t]he things we call ‘technologies’ are ways of building order in our world.” The situation in the People’s Republic of China is extreme; but it is not unique. It brings to light, in sharpened form, a tension between algorithmic functionality and legal accountability which has not been resolved by any jurisdiction yet.

The paper Rule by Algorithm was published in the Columbia Journal of Transnational Law 63, (2025). Michelle Miao is an Associate Professor at the Faculty of Law, the Chinese University of Hong Kong, and was a Fellow of the Center for Advanced Study in the Behavioral Sciences at Stanford University when she wrote this article.

The statements made in this blog post do not reflect the views of the blog editor.

Algorithmic Control and Judicial Conformity: How China’s Smart Courts Redefine Justice

16. March 2026

By Jiajun Luo

When China first introduced Smart Court reforms in 2010s, the official narrative emphasized efficiency, transparency, and modernization. Court workloads were exploding, the public wanted faster and fairer judgments, and technology promised a way forward. Judicial AI systems could streamline case management, assist judges in drafting opinions, and even suggest appropriate sentences.

But beneath this techno-optimistic façade lies a deeper story: Smart Courts are not just digital upgrades. They are political projects. This research seeks to examine how these reforms, presented as neutral innovations, actually reshape the relationship between courts, judges, and the Party-state. Far from empowering frontline judges, they are transforming courts into algorithmically managed bureaucracies—where uniformity and control increasingly trump deliberation and discretion.

China’s courts face a real dilemma: they must process millions of cases each year, yet judicial discretion—the capacity of judges to interpret law flexibly—has long been seen by the Party as a potential political risk. Under President Xi Jinping, judicial reforms have sought to address both challenges simultaneously: accelerating case processing while tightening constraints on judicial discretion, not only in politically sensitive matters but also in routine adjudication. One provincial High Court described this shift vividly: “We are moving from person-watching-person to algorithm-watching-person.” Algorithms now stand between judges and their decisions, monitoring not only what they rule, but also how they reason.

My fieldwork and document analysis identify three main technologies driving this transformation. The first is mandatory similar-case retrieval, which requires judges to search for algorithmically selected “similar cases” before issuing rulings. Systems flag deviations from these precedents, nudging judges toward doctrinal conformity. What counts as a “similar case” is often defined not by law but by internal court authorities. Second, case deviation alerts trigger warnings when a draft judgment diverges from prior rulings or administrative guidelines. Some platforms even propose “corrections” automatically. Judges must justify deviations to court leadership before proceeding, creating a presumption against independent reasoning. Third, AI-generated judgments are still experimental but growing rapidly. These systems can now draft entire judicial opinions—from facts to legal reasoning to final verdicts—based on pre-set templates. In one Jiangsu court, judges described the platform as offering “multiple-choice justice,” with little room to insert their own views before finalization.

On paper, Smart Courts promise greater access to justice. Case filing is now automated; citizens can submit complaints online; judgments are published in massive digital databases. Yet in practice, access to justice is becoming access to automation. Litigants report that complex or unconventional claims risk being rejected by keyword-based filing systems. For judges, algorithmic deviation alerts discourage nuanced reasoning. Appeals are becoming less meaningful because decisions already carry the weight of algorithmic legitimacy.

From the court user perspectives, three particular consequences stand out: procedural rights such as the right to be heard or to appeal shrink as decisions are pre-shaped by internal systems rather than open deliberation; legal reasoning becomes standardized, as judges start from algorithmic “recommended outcomes” and then search for precedents to justify them rather than reasoning from facts and principles; and substantive fairness suffers as unique or emotionally complex cases are forced into one-size-fits-all templates.

The rise of Smart Courts marks a turning point in China’s legal system. What began as a pragmatic response to overwhelming caseloads has evolved into a project that reshapes the very logic of judicial decision-making. These reforms promise efficiency and consistency, but they also redefine the role of courts and judges in ways that deserve close attention. In a closed political system like China, the stakes are especially high. Judicial independence has long been off the table, but now judicial autonomy is shrinking as algorithms embed uniformity and administrative oversight into routine decision-making. Legal reasoning increasingly risks taking a back seat to compliance with algorithmic and political mandates. Over time, public trust in the courts may erode if justice comes to feel automated, standardized, and disconnected from the complexities of real-world disputes.

Technology now reinforces both legalism and state control, often at the expense of judicial autonomy. This trajectory raises pressing questions about the future of law and justice in China. Will legal reforms continue to privilege speed and uniformity over reasoned judgment and fairness? And how much discretion will judges retain in a system governed by algorithms, administrative priorities and Party imperatives?

Jiajun Luo is a Hauser Postdoctoral Global Fellow at NYU School of Law. His research explores public law, dispute resolution, and authoritarian governance. He can be reached at jiajunlok@gmail.com.

China’s Export Control of Dual Use Items: From Multilateralism to Extraterritoriality

27. February 2026

A new paper by Alexandr Svetlicinii and Jing Wen

Historically, the People’s Republic of China’s (PRC) approach to export controls has been shaped by its international obligations under multilateral non-proliferation regimes. Even amid geopolitical tensions, China’s adherence to multilateralism in this domain was reinforced in its 2021 White Paper on export controls. It was not until geopolitical tensions further escalated and various forms of economic restrictions proliferated that the PRC moved to modernise its export controls by adopting the 2020 Export Control Law.

China’s export control system has undergone systematic restructuring. The items subject to export controls were previously scattered across various regulatory instruments and were then consolidated into a single, unified control list. On the institutional side, a centralized governance hub was established, with the State Council and the Central Military Commission overseeing policy formulation, the Ministry of Commerce (MOFCOM) carrying out the enforcement, and the Ministry of Foreign Affairs, the Ministry of National Defense, and other agencies aiding in enforcement, thus ensuring a more centralized and systematic approach to export control governance. Additionally, the reformed regulatory framework incorporates the “watch list” and “control list,” which operate in synergy with the “unreliable entity list” administered by MOFCOM.

Strategically, China developed a distinctive “few but precise” approach to designating items subject to export control. In stark contrast to the lengthy lists of controlled items maintained by the U.S. Department of Commerce and the European Union, China’s unified control list contains about 700 items, focusing on strategic products and associated technologies. These include critical minerals such as gallium, germanium, tungsten, and indium, as well as several categories of rare earth elements, including samarium, gadolinium, and terbium, which are crucial for high-end sectors such as semiconductors, new energy, and advanced weapons manufacturing. By leveraging its control over these resources, China strengthened its bargaining power in global supply chains and bolstered its retaliation potential.

Furthermore, the PRC maintains a flexible mechanism that allows for transitioning between temporary and permanent controls, which is conducive to both responding to geopolitical shifts and stabilising trade flows. For example, graphite products transitioned from temporary control in 2006 to permanent control focused on high-end graphite materials in 2023, with exemptions for ordinary industrial graphite. This approach ensures targeted protection of strategic interests while preserving space for regular trade.

As critical technologies and strategic resources from China are integrated into global production networks, it has become evident that controlling them solely within the PRC’s territory is insufficient to achieve regulatory objectives. As a result, the 2024 Regulation on the Export Control of Dual-use Items provides for the extraterritorial application of PRC export control rules in relation to foreign-made products that contain PRC-origin dual-use items, goods produced abroad using PRC technology, and the transhipment of PRC-origin dual-use items. This framework draws inspiration from the U.S. export controls, such as the minimum content rules and the Foreign Direct Product Rules. However, it is not a mere replication but adapted to fit China’s non-proliferation obligations and national security interests. For example, foreign products containing components originating in the PRC, even if assembled abroad, must comply with PRC export control regulations if they contain controlled items. Similarly, control over PRC technology used abroad focuses on strategic technologies such as high-performance drones and semiconductors. The export control regulations further introduce defensive measures against the extraterritorial application of foreign export controls. Chinese companies are prohibited from cooperating with inspections launched by foreign governments without prior approval from the MOFCOM.

The transformation of China’s dual-use export control regime is not just an update of domestic regulations. It demonstrates that the PRC reshapes export controls with a substantial effect on global supply chains. The asymmetries and overlaps between China’s export control systems and those of other countries pose substantial compliance challenges for multinational corporations. The lack of clear extraterritorial applicability thresholds and the inherent flexibility of national security criteria further complicate matters, creating commercial and compliance risks for enterprises as they make decisions about supply chain configurations, technological cooperation, and other business strategies.

The full paper, titled China’s Export Control of Dual Use Items: From Multilateralism to Extraterritoriality, is published in the Global Trade and Customs Journal, vol. 21/2026. This research was supported by the University of Macau Multi-Year Research Grant MYRG-GRG2023-00209-FLL “Regulation of the Dual Use Goods in International Economic Law”. Alexandr Svetlicinii is an Associate Professor of Global Legal Studies at the University of Macau, Faculty of Law. He can be contacted at alexandrs@um.edu.mo. Jing Wen is a Master of Laws candidate at the University of Macau, Faculty of Law.

When Algorithms Meet Justice: A Deep Dive into AI-assisted Criminal Proceedings in China

15. January 2026

By Wanqiang Wu and Xifen Lin

Picture a prosecutor in Shanghai opening their computer to review a theft case. Instead of manually searching through thousands of precedents, they turn to an AI system that instantly analyzes the evidence and recommends similar cases from a database of millions. Within minutes, the system suggests whether to detain the suspect and even predicts the likely sentence. This isn’t science fiction anymore, it is the reality of China’s “206” system, arguably one of the most ambitious experiments in AI-assisted criminal justice worldwide.

The Rise of Smart Justice in China

Since 2016, China rapidly embraced what it calls “Smart Justice” (智慧司法), integrating artificial intelligence throughout its legal system. The Shanghai “206” system represents one frontline pioneer of this transformation. Unlike AI applications in “Western” courts that focus on specific tasks like risk assessment, China’s system attempts something far more comprehensive: it assists judges and prosecutors across the entire criminal process, from pre-trial detention decisions to sentencing recommendations.

What makes this particularly fascinating is not just the technology itself, but the institutional context. China’s judicial system, operating without the constraints of judicial review traditions found in many “Western” courts, has adopted AI technologies with remarkable speed and minimal resistance. The AI-assisted system now processes hundreds and thousands of cases annually, with the “206” system expected to be utilized in all criminal cases in Shanghai.

How Does It Actually Work?

The “206” system’s core feature is its Similar Case Recommendation function, which operates like a sophisticated legal search engine on steroids. When a prosecutor inputs case details, the system uses deep learning algorithms trained on past verdicts to identify patterns and recommend outcomes. It considers over 50 variables, from the suspect’s employment history to whether victims have forgiven the accused, to generate sentence recommendations.

The system does not just help with individual decisions. For example, in plea leniency cases (which now account for 87% of criminal cases in China), prosecutors may display the AI’s predictions to negotiate with defendants. “Look,” they might say, “based on our AI system, you’re likely facing 3-5 years of imprisonment. If you plead guilty now, we can recommend the lower end according to the law.” The system has changed the dynamics of criminal justice negotiations.

The Good, The Bad, and The Algorithmic

Our research, building on exclusive data from Shanghai’s procuratorate, reveals a complex picture. On the positive side, the data suggest genuine improvements: cases processed with AI assistance took 23% less time to complete, and sentencing recommendations made with AI support were accepted by judges 75.8% of the time, whereas they only accepted 65.6% of the sentencing recommendations without AI support. The system appears to reduce arbitrary detention and increase consistency in sentencing.

However, three critical concerns emerged based on our fieldwork. First, the anchoring effect: when prosecutors see the AI’s recommendation first, it becomes incredibly difficult for them to deviate, even when case specifics might warrant a deviation from the recommended sentence. Once the prosecutor sees the recommended number on the screen, she or he ascribes a sometimes unwarranted authority to it.

Second, accountability avoidance: the system’s complexity creates a perfect excuse for passing the buck should errors occur in sentencing. If something goes wrong, was it the algorithm’s fault? The fault of the system developer? The fault of the prosecutor or judge who relied on the system? This diffusion of responsibility poses serious challenges to China’s judicial accountability reforms.

Third, and perhaps most troubling, is the compression of the rights to a proper defense. Defense lawyers have no access to the system, cannot challenge its algorithms, and often do not even know that it is being used. While prosecutors wield sophisticated AI tools funded by public money, defendants and their lawyers are left in the dark. This to some extent amplifies the already existing problems of an unequal playing field in criminal justice.

Lessons for the Global Legal Community

What can the rest of the world learn from China’s bold experiment? First, procedural design matters more than technological sophistication. Our research suggests that many of the risks associated with AI in criminal justice are not inherent to the technology but arise from how it is implemented. Simple procedural safeguards, like requiring judges to form initial opinions before consulting AI, or ensuring that the defense has access to algorithmic tools, could mitigate many concerns.

Second, transparency isn’t optional. The closed nature of China’s system, where algorithms operate as black boxes hidden behind claims of trade secrets of technology providing companies, undermines procedural justice. Any jurisdiction considering adopting AI for prosecution or adjudication work must grapple with balancing technological innovation with fundamental legal principles like the right to a fair defense.

Finally, China’s experience confirms a widely observed paradox: AI systems designed to reduce human bias and increase consistency may actually entrench existing patterns of injustice if they are trained on historical data reflecting those very biases. The algorithm does not innovate; it replicates and amplifies patterns that have previously manifested.

Looking Ahead

As courts worldwide grapple with backlogs and inconsistencies, China’s aggressive adoption of AI offers both inspiration and cautionary tales. The technology clearly has potential to improve efficiency and consistency in criminal justice. But our research suggests that without careful attention to procedural safeguards, transparency, and equal access, AI applications risk creating a two-tiered system of justice where algorithmic efficiency trumps fundamental fairness. The question is not whether AI will transform criminal justice, but whether we can design systems that harness AI’s benefits while preserving the procedural protections that define justice itself.

The full article titled Access to technology, access to justice: China’s artificial intelligence application in criminal proceedings can be accessed here.

Wanqiang (Aiden) Wu is a Yat-sen Postdoctoral Fellow at Sun Yat-sen University Law School. He received his Ph.D. in Criminal Procedure Law (Cum Laude) from Shanghai Jiao Tong University in 2025. His research focuses on criminal procedure, empirical legal studies, and the intersection of technology and criminal justice in China. He has published extensively on China’s procuratorial system and judicial reforms in leading journals including Modern China, Hong Kong Law Journal, and International Journal of Law, Crime and Justice. Contact him via email.

Xifen Lin is Professor of Law and Vice Dean at KoGuan School of Law, Shanghai Jiao Tong University. Professor Lin is a leading scholar in Chinese criminal procedure and judicial reform, with particular expertise in prosecutorial systems and empirical legal studies. Contact him via email.

Big Data Cannot Answer All Questions About Chinese Courts

4. December 2025

By Benjamin Minhao Chen

Imagine a researcher seeks to answer a fundamental question of legal fairness: Do better-resourced parties (the “haves”) achieve more favorable outcomes in Chinese courts simply because of their socio-economic status?

A naïve answer is to compare the win rates of well-resourced and less-resourced parties in litigation. But the researcher might quickly realize that well-resourced parties are likely to be represented by counsel whereas the less-resourced are likely to be self-represented. Perhaps it is the quality of legal representation that influences how judges rule, not the status of the parties. On this account, the “haves” do better than the “have nots” because they have, among other things, superior legal representation.

A solution might be to control (or adjust) for legal representation. It is common for statistical studies to control for an explanatory variable to isolate the effect of the variable of interest on outcomes. Controlling for legal representation here means, essentially, that the “haves” who are represented by counsel are compared to the “have nots” who are represented by counsel. These comparisons seem to make sense: if the researcher finds that represented, well-resourced parties still win more than represented, less-resourced parties, that indicates that status is driving the observed pattern of outcomes.

This approach can result in inaccurate inferences being drawn due to collider bias. In causal modeling, a collider is a variable that is a common effect of two other factors. A classic, non-legal example is the car that fails to start. This can be caused by a dead battery or an empty gas tank.[1] These two causes are independent; one doesn’t cause the other. Suppose you only look at cars that have broken down, i.e. you condition on the collider “car fails to start”. Then a correlation appears between the two causes: if you know a broken-down car has gas, you can be confident the battery is dead. But of course, the car having gas is not itself a cause of the battery being dead.

So how does collider bias apply to our example of the “haves” and “have nots” in Chinese courts? There is suggestive evidence that Chinese litigants are more likely to seek professional advice in harder, more doubtful, cases.[2] So, if a well-resourced party has a lawyer, it could be because they can easily afford one or because the merits of their claim are doubtful. If a less-resourced party has a lawyer, it is probably because the merits of their claim are doubtful. If a researcher controls for legal representation, she is comparing the “haves” who are represented by counsel are compared to the “have nots” who are represented by counsel. But the “haves” who are represented by counsel have, on average, stronger cases on the merits than the “have nots” who have counsel, and they might therefore prevail in more cases. The researcher may conclude from her controlled comparison that the “haves” come out ahead more often in Chinese courts because of who they are. But this inference is not necessarily warranted by the evidence.

This example demonstrates that big data does not absolve legal scholars from thinking through the causal relationships between the variables in their analyses. Indeed, qualitative and sociological methods can produce valuable domain knowledge for distinguishing plausible relationships from spurious ones. A multi-disciplinary paradigm remains critical to studying the Chinese legal system, even in the era of artificial intelligence.

The full article, titled Data Still Needs Theory: Collider Bias in Empirical Legal Research, co-authored by Xiaohan Yin, is available here. Benjamin Minhao Chen is Associate Professor and Director of the Law and Technology Centre at the University of Hong Kong Faculty of Law. You can contact him via email.

[1] Judea Pearl, Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference (San Francisco: Morgan Kaufmann 1988).

[2] Wenzheng Mao and Shitong Qiao, ‘Legal Doctrine and Judicial Review of Eminent Domain in China’ (2021) 46 Law & Social Inquiry 826; Yali Peng and Jinhua Cheng, ‘Ethnic Disparity in Chinese Theft Sentencing’ (2022) 22 China Review 47

The Iron Fist in a Digital Glove: How the CCP Manages Its Members with Apps

24. November 2025

By Ningjie Zhu

In our daily lives, there’s an app for almost everything—ordering food, tracking fitness, managing work. But what if there was an app for political loyalty? For the Chinese Communist Party (CCP), this isn’t a hypothetical question. It’s a core part of its strategy for governing in the 21st century. Through a massive initiative known as “Intelligent Party Building” (智慧党建), the CCP is rolling out digital platforms to manage, educate, and discipline its 95 million members.

My research explores how this digital push is reshaping the very nature of power within the Party. It is not just about efficiency; it is about deepening control. The Party aims to achieve two seemingly contradictory goals at once: build a more modern, responsive bureaucracy while simultaneously reinforcing absolute top-down authority. This creates a new, powerful dynamic of internal governance in the digital age.

The Two Faces of State Power: An Iron Fist and a Nervous System

Despotic Power: This is the “iron fist”—the state’s ability to make decisions and issue commands without negotiation. It’s top-down, coercive, and absolute. Think of it as power over society.

Infrastructural Power: This is the state’s “nervous system”—its actual capacity to penetrate society and carry out its plans. It relies on logistics, institutions, and technology to get things done. It is power through society.

Historically, regimes were often strong in one of the types of power but weak in the other. For example, an ancient empire might have immense despotic power (the emperor’s word is law) but weak infrastructural power (it is hard to actually collect taxes in a remote province). Modern democracies often have strong infrastructural power (efficient services) but weak despotic power (leaders are constrained by law and public opinion).

The CCP under Xi Jinping, however, is trying to maximize both. “Intelligent Party Building” is a prime example of this ambition, using digital tools to build a highly efficient administrative nervous system that also serves a powerful iron fist.

Inside the “Party-Building Cloud Platforms”

So, what do these apps and platforms actually do? While they vary locally, they generally focus on four areas:

Digital Dossiers and Performance Tracking: The traditional paper dossier (dossier), a lifelong file tracking a person’s political behavior, is going digital. These platforms create a permanent, tamper-proof record of a member’s participation in Party activities. Poor performance can trigger warnings, public criticism, or even negative entries in their file that could impact their career.

Ideological Education on Demand: The platforms deliver a constant stream of ideological content. This includes live-streamed lectures, articles from state media, and materials from official education campaigns. To ensure engagement, members are often tested with gamified quizzes where they can earn points and compete on leaderboards.

Gamifying Loyalty: The platforms borrow heavily from popular apps to keep users engaged. For instance, the famous “Study the Great Nation” (学习强国) app rewards users with points for reading articles about Xi Jinping or watching videos of his speeches. This subtly transforms political indoctrination into a daily habit, much like checking social media.

Digitizing Bureaucracy: Beyond ideology, these platforms are also practical office tools. They handle routine tasks like paying Party fees, managing announcements, and approving leave requests. By integrating with daily work, they ensure the Party’s presence is not just a separate political activity but is embedded into the fabric of everyday professional life.

A Double-Edged Sword of Digital Control

The true innovation here is not the technology itself—many of these features mimic existing corporate collaboration tools like Ding Talk or WeChat Work. The innovation lies in its dual-use application for political control.

The very same feature that enhances infrastructural power (e.g., efficiently collecting data to provide personalized educational content) is used to wield despotic power (e.g., using that same data to monitor and punish members for insufficient engagement). A tool designed for convenience also becomes a tool for surveillance. A gamified quiz that makes learning “fun” is also a mechanism for ideological enforcement.

Ultimately, the CCP’s push for “Intelligent Party Building” reveals a fundamental tension. While it seeks to modernize its internal management to become more efficient and responsive, it remains unwilling to give up the coercive, top-down control that defines its Leninist roots. The result is a system where the iron fist is now wearing a digital glove, able to reach further and grip tighter than ever before.

Ningjie Zhu is a researcher at the Center for Advanced Security, Strategic and Integration Studies (CASSIS) at the University of Bonn. You can reach him at nzhu[at]uni-bonn.de.

Smart Court, Smarter Party: A Necessary but Incomplete Interpenetration

17. November 2025

By Larry Catá Backer

The idea of “smart courts” has become a globally shared ambition. From the European Commission’s “Digitalisation of Justice” toolbox to the U.S. judiciary’s call for more efficient digital infrastructures, and with examples emerging from Egypt to Tanzania, the movement to enhance the judicial function with technology is a planetary phenomenon. Yet, as I explored during the workshop Smart Courts in Comparative Perspective at the University of Cologne this July, no national project invites as much theoretical scrutiny—and perhaps as much political resonance—as China’s.

China has given this movement a name—or at least rebranded the product of the alignment of technology and the courts: Smart courts, zhihui fayuan 智慧法院. The name suggests an alignment on the ground that is both linguistic and textual in the operational spaces of courts. Over the last decade or so, and through its Supreme People’s Court, Chinese officials have led a national effort to modernize the judicial system through the use of emerging technologies. Like other modernization pathways elsewhere, the goals include enhancing access to justice and ensuring that access provides pathways toward just outcomes. Since December 2024, these efforts also include an artificial intelligence platform to help judges improve work efficiency.

But names sometimes are a distraction. And that appeared to be the case with Chinese smart courts. The name became a vessel into which people could pour their larger fears about the transformations they feared most—that the people would no longer be their own masters but would serve technology even as technology appeared to serve them. It is no surprise, then, as Susan Finder relates in her examination of the Supreme People’s Court 2024 Work Report to the National People’s Congress, that the term “smart courts” appears to have been dropped.

Nevertheless, “smart courts” have become not merely a symbol of digital reform but a mirror reflecting deeper ideological and systemic transformations. What appears at first to be a techno-administrative modernization effort quickly reveals itself to be an exercise in high-stakes governance theory. The central question I pursued: Can a digitally advanced judiciary maintain alignment with a ruling party that is not itself digitally transformed? In other words, can a smart court operate effectively without a smart Party?

From Robot Courts to Zombie States?

The study of “smart” or “intelligent,” or “wise” courts can be approached from a large number of perspectives. I start from the ordering premise that these “smart” courts can be understood as an object, and also as a symbol or signified conception, and lastly as the set of objects and behaviors that produces its own meaning through its own dialectical phenomenology—that being by doing. This amalgamation of objects and symbols is a matter central to the continued evolution, in human society, of the notion and practice of judging, and of institutions of judging to which it is both attached and to which it lends meaning. But an object and symbol of what?

The term “smart court” evokes both utopian promise and dystopian anxiety. While the ambition of the People’s Republic of China has been to develop courts that are faster, more accessible, and more consistent, the term has also sparked deeper fears—especially outside China—of robot judges, automated justice, and dehumanized legality. This isn’t merely science fiction. Rather, as I suggested during my talk, these fears can be metaphorically grouped into a three-course cautionary tale.

First, courts risk being consumed by the very technology meant to assist them. Their core identity shifts from a site of judgment to a platform for automated processing. Second, courts may begin to consume their stakeholders—litigants, judges, and lawyers—by reducing them to data points in algorithmic workflows. Third, courts may consume themselves, becoming mechanisms of predictive governance rather than instruments of legal deliberation.

Such risks are not unique to China. But within China’s governance model, they raise particularly intense contradictions—especially the one between technology-led modernization and Marxist-Leninist political control.

Semantics Matter: What Is “Smart”?

Much of the misunderstanding about smart courts, I argue, stems from the loaded semantics of “smartness” itself. In English, “smart” blends quick wit, technological capacity, and sometimes pain (its etymology rooted in “to sting”). In Chinese, however, the distinction is sharper. Zhìnéng (智能) points to technical capability—what we associate with AI and data-driven systems. Zhìhuì (智慧), by contrast, suggests discernment, judgment, and wisdom.

This duality—between instrumental intelligence and human wisdom—is crucial. Smart courts, if they are to serve justice rather than mere efficiency, must retain a core of hui: the human capacity to judge wisely. In the Chinese political imagination, this is ideally embodied by the judge and the collective judiciary. But what happens when the source of wisdom—traditionally human—is threatened by ever-smarter systems?

Tech as Instrument, or as Actor?

China’s digital judiciary remains in a transitional phase—digitisation more than full digitalisation. The emphasis is still on improving efficiency: filing systems, access to records, online hearings. Yet, the horizon is shifting. Predictive analytics, caselaw modeling, and AI-assisted adjudication point to an emerging reality where tech not only facilitates justice but begins to shape its substance.

This introduces a profound conceptual tension. As technology moves from being “smart” (responsive and efficient) to potentially “wise” (autonomous and analytical), it also shifts from being a tool to being an actor. This challenges long-standing assumptions about who—or what—gets to decide within a legal system.

The Smart Court Needs a Smart Party

This transformation becomes most consequential in China, where courts are not isolated institutions but deeply embedded within a Party-led governance model. The CPC is not just a political overseer but the ideological architect of the judiciary’s function. Here, smart courts demand something deeper: a smart Party.

By “smart,” I mean a Party apparatus that itself incorporates digital technologies not only in surveillance and administration, but in its very processes of leadership, assessment, and ideological guidance. Without such a transformation, an asymmetry emerges: the courts grow in techno-capacity, while the Party lags in digital adaptability. That gap threatens to destabilize the very premise of Party-led governance.

Rethinking Interpenetration: Court and Party

Chinese governance is structured around interpenetration—the mutual embeddedness of Party and State institutions. Historically, this interpenetration has been managed through personal-bureaucratic forms: overlapping roles, dual appointments, and ideological campaigns. In the digital age, however, interpenetration is reconfigured through data flows, predictive modeling, and feedback loops.

The smart court, then, is not just a site of dispute resolution but a generator of political data—inputs and outputs that reflect the health of Party ideology and administrative discipline. To oversee such a system, the Party must itself become a digitally competent, analytically capable, and ideologically precise actor.

This is no small task. It means building a digitally-enhanced Party apparatus that can assess court behavior, monitor ideological conformity, and even model the likely impact of judicial decisions—all without becoming a mere appendage of the technologies it deploys.

The smart court exemplifies both the achievements and the contradictions of China’s New Era. On the one hand, it reflects the success of socialist modernization: the integration of productivity-enhancing technologies into governance. On the other hand, it surfaces a contradiction between human-led ideological guidance and machine-augmented decision-making. Two key contradictions define the current moment. First, the contradiction between the leadership of the Party and its capacity to lead in a tech-driven environment. Second, the contradiction between technology as instrument and technology as autonomous force. Both must be addressed if the CPC is to retain its position as the core of the political-economic order.

Ultimately, one must come to understand, or at least consider the plausibility, of a principle that under New Era Chinese Marxist-Leninism, the state apparatus can only be as “smart,” intelligent” and “wise” as it is in the capacity and operations of the Party to do likewise. In the presence of asymmetry two fundamental contradictions must be addressed. The first is the contradiction between the leadership of the Party and its capacity to lead. The second is between the techno-instruments through which Party capacity is undertaken and the ability of the Party apparatus to steer, guide, assess, control and utilize these instruments in the performance of its own duties and responsibilities. The fundamental issue of instrumentalization and capacity remains undisturbed—the more autonomous the technology, the greater the risk that the relationship between instrument and its wielders will be reversed, at least in part. In the absence of a capacity to understand and manage those contradictions, either organs better capacitated to wield techno-instrumentalized applications and processes will drive human collective systems, or human collective systems may become an instrument through which techno-wisdom intelligence may realize its own vision for techno-human perfectibility.

Implications Beyond China

While my analysis focuses on the Chinese context, the underlying challenges are global. Whether in Europe, the U.S., or elsewhere, legal systems face similar dilemmas: How to preserve human judgment in algorithmic environments? How to ensure accountability when decisions are guided by machine learning? How to maintain institutional integrity when data becomes both input and output? China’s smart court project offers a provocative case study. It forces us to confront not only what technology can do for justice, but what it might do to justice—and who, in the end, will be wise enough to decide.

The full contribution is available here. Larry Catá Backer is the W. Richard and Mary Eshelman Faculty Scholar and Professor of Law and International Affairs at Pennsylvania State University. His work focuses on Chinese governance, transnational law, and political theory.

Children’s Tort Liability and Evolution of Chinese Tort Law in the Post-Codification Era

15. August 2025

A new paper by Yuanshi Bu

How do ordinary Chinese people think about justice? Tort law may provide an excellent lens through which to explore this question. Within tort law, the issue of children’s tort liability stands out as particularly illuminating, as it highlights the significant challenges families and schools face in raising, caring for, and educating children in China. Four years after the Chinese Civil Code (CCC) came into effect, more than 10,000 court judgments related to children’s tort liability have already been made available online. Thus, it came as no surprise that the Supreme People’s Court’s first judicial interpretation of the Tort Liability Book of the CCC devoted half of its content to children’s tort liability. After all, the CCC has codified six norms in this area, with three addressing children as tortfeasors and three pertaining to children as victims.

However, Chinese legal scholars often criticize the domestic approach to handling children’s tort liability and have made various proposals for improvement based on foreign models. To understand why these proposals have not been adopted by the judiciary, and to uncover the logic behind the development of Chinese tort law, I examined around 400 relevant court decisions. In recent years, big data analysis of Chinese tort law is becoming increasingly popular. However, given my interest in factual patterns and judicial reasoning, I found it more meaningful to review actual court judgments directly, despite the considerable amount of time associated with it. The judgments I examined revealed the following characteristics of the legal regime governing children’s tort liability:

First, the Chinese regime of children’s tort liability is characterized by a liability union in respect of a minor and his or her parents. This means that if children cause harm to others, both they and their parents are held liable. However, the liability of the child wrongdoer and that of their parents is considered one and the same. This stands in stark contrast to other jurisdictions such as the German one, where parental liability is separate from that of their children. For example, unlike under Chinese law, under German law parents are not liable if they have properly educated their children not to play with fire, and the children nevertheless did so and burned down a house. In addition, in the Chinese liability regime, even toddlers can be considered capable of committing tortious acts. For instance, liability was imposed on the parents of a 22-month-old toddler who ran over the victim’s foot with a buggy. In this sense, tort liability applies to children without a minimum age threshold. Parental liability is incurred automatically, regardless of whether the child is under the parents’ care, even when the harm occurs in a boarding school. A child wrongdoer’s family can be released from liability only if it can show that the harm was entirely caused by the victim or a third party. Under the current legal regime, grandparents in China, who often serve as primary caregivers for young children without emuneration, also face considerable liability risks.

Second, the harsh liability regime is mitigated by the fact that Chinese courts rarely assign 100% liability to child wrongdoers and their parents. Child victims are often held partially liable for their own injuries, regardless of their age. In one case, a nine-year-old child was injured by two other children (aged nine and eleven) who had built an improvised swing from strips of artificial grass on a mobile soccer goal at school, causing it to tip over. Although the victim had done nothing more than stand nearby, he was found 20 % at fault. In the eyes of Chinese judges, being at the wrong place at the wrong time already indicate a failure of the child or the parents.

Third, the above-mentioned finding basically holds true also with regards to harms occurring in schools. While many relevant cases involve minors colliding and losing their (milk) teeth, serious incidents also occur in which victims have lost their eyesight or even their lives. To reduce their exposure to liability, many schools, particularly elementary schools, do not allow pupils to leave the classroom during breaks, except to use the toilet. Additionally, an increasing number of school services, such as transportation and lunch, are being outsourced.

One might ask why parents and schools do not simply take out liability insurance to solve the problem. The answer is that ordinary Chinese still tend to try their luck, at least up to now. In most cases, neither the victim nor the tortfeasor has liability insurance. Even worse, medical coverage is particularly precarious for children: 60.44 % of the treatment costs are reportedly born by the families. This is why people sometimes litigate over relatively minor injuries. Confronted with high vulnerability of the child wrongdoer and child victim in coping with accidents, courts in China often have no choice other than distributing losses among all involved parties to avoid social discontent. In this sense, tort law in China is still loyal to the conventional primary purpose of compensating victims.

Lastly, through reading the judgments, a discrepancy between legal scholarship and the judiciary regarding the perception of what constitutes deficiencies in tort law has become apparent. In fact, it is still extremely rare for Chinese scholars to consider court rulings when researching this topic. Instead, emphasis is placed on comparative law. I believe that if the goal of Chinese legal scholarship is to build an autonomous knowledge system and thereby provide guidance to the legislature and judiciary, more attention must be paid to court practice in China and the rationale behind adjudication patterns. So far, due to the social, economic, and political context in which Chinese tort law operates, it has followed a rather need-based and piecemeal path.

The full paper, titled “Evolution of Chinese Tort Law in the Post-Codification Era: A Comparative Study Based on Children’s Tort Liability”, is published in the German Journal of Chinese Law (in English), vol. 2/2025. Yuanshi Bu is a Professor and the Director of the Chair of International Economic Law with a Focus on East Asia at the University of Freiburg. She can be contacted at asien@jura.uni-freiburg.de.

Gendered Costs of Human Capital Upgrading: The Case of Garment Workers in China

24. July 2025

A new paper by Yiran Zhang

“Student companion mothers” in their workshop. Photo by Yiran Zhang

As “the factory of the world,” China’s economy was once driven by a massive cheap-commodity manufacturing sector that relied on a cheap and informal workforce of internal migrants. Since the late 2000s, the People’s Republic of China has embraced a new developmental regime that focused on “upgrading” its manufacturing economy through industrial policies and enhancing the quality of its workforce through labor, welfare, and population law reforms, diverting the Chinese economy away from a cheap-labor regime. The industrial upgrading side, once known as “Made in China 2025,” invested in high-tech and more skill-demanding industries. On the “social upgrading” front, the state introduced more labor regulations, made significant efforts to close the urban-migrant social welfare gap, and promoted the stability of migrant families, aiming to transform its manufacturing workforce into one more formal and urbanized. Nevertheless, I find that the heavy dependence of this upgrading reform on parents’ role in the human capital investment of the next generation’s workforce—lacking sufficient public support—led to a downgrading of today’s manufacturing jobs, especially for women, and a new dynamic of gendered supply-chain subordination.

My finding draws from my 2018-19 fieldwork in the Yangtze River Delta that traced the reconfiguring of labor and migration norms across the hometown and worksite of a group of garment workers. In my field work, I observed a group of middle-aged, female industrial workers who had been laboring in coastal factories alongside their husbands for decades, recently returned to their inland hometowns, precisely at a time when labor and social reforms “upgraded” their old factory jobs into higher-quality employment. This trend of reverse migration was primarily motivated by an emerging community norm that may be translated as “student companion mothers (陪读妈妈)” in the region. As mothers of teenage children, women increasingly felt the community expectation that they should return to their hometown to accompany their children and support them as they entered middle or high school. In China, the household registration system (戶口) creates obstacles for students to go to school outside their permanent domicile. As parents seek work opportunities in coastal areas, they often have to leave their children in the care of their grandparents or other family members. Women increasingly decide to return to their hometowns, consciously prioritizing the demands of their children’s education over job and micro-entrepreneur opportunities at the East coast. “Student companion mothers” often live with their children in group housing next to the school and perform care as meticulously as possible, hoping to generate more study time for their children and thus boost their performance in crucial exams. The norm of intense parenting, which started among urban families in the early 2000s, has reached and adapted to migrant families.

This change in family norms also restructured the regional garment industry. The en masse return of female migrant workers gave rise to a new wave of home-based “mothers’ workshops” around inland schools that subcontracted sewing and other labor-intensive tasks from coastal factories. Owned and staffed by student companion mothers, these workshops form a new segment of precarious work— work that is poorly remunerated, low in productivity and contingent, and yet highly flexible as schedules are tailored to fit care work mothers perform for their children in school. The female workers accept what is less than one-third of their old wages because returning to their hometowns transformed their roles within their household from an equal wage earner to a caregiver. The contingency of this workforce also constrained these inland workshops’ bargaining power in the supply chain, creating feminized enclaves in the manufacturing economy that evade both the labor reform and the state’s industrial upgrading efforts.

My case study of garment workers in China raises three interconnected implications for labor and development law and policies. First, precarious work and women’s unpaid social reproductive labor is structurally integrated. Both the institutional arrangements and normative standards of social reproduction condition female workers’ participation in the paid workforce. Thus, mere legal reforms to paid jobs may not be adequate to raise women’s labor conditions. Second, value-chain-upgrading reforms—the primary policy prescription promoted by the mainstream global value chain literature—have a potential gender impact. As one key element in value-chain upgrading strategies, namely human capital accumulation, inevitably relies on highly gendered social institutions such as the household, it is worth examining how such development strategies distribute the costs and gains of upgrading in and beyond the value chain. Third, it remains of crucial importance to understand population and family policies as an integral part of the People’s Republic of China’s development policies. Such a more comprehensive understanding of development policies not only captures a more holistic picture of the state-society relationship, but also reveals that expectations a policy regime imposes on actors in its economy are competing and sometimes contradictory.

The blog post is based on a pair of articles: the sociological analysis, entitled “The Paradox of Upgrading: Standards of Social Reproduction and the Gendered Precarization of Garment Work in China,” appeared in Critical Sociology, and the policy analysis, entitled “Gender, Value-Chain Upgrading, and The Costs of Human Capital: The Case of a Garment Supply Chain in China,” appeared in Cornell International Law Journal vol 57 (2024). Yiran Zhang is the Proskauer Assistant Professor of Employment and Labor Law at Cornell University Industrial and Labor Relations (ILR) School and an Associate Faculty Member of the Cornell Law School. She writes about care, work, gender, and the law in the US and China. Her scholarship is available on SSRN.