Fa Fa [2026] No. 10: How the Supreme People's Court Explains Its First Adjudication Rules for AI Disputes

Editor’s Note — DCC.

On 7 September 2026 the Supreme People’s Court (最高人民法院, SPC) held a press conference to release the Opinions of the Supreme People’s Court on Trying Cases Involving Artificial Intelligence Disputes in Accordance with Law (最高人民法院关于依法审理涉人工智能纠纷案件的意见), document number Fa Fa [2026] No. 10 (法发〔2026〕10号). Vice President Tao Kaiyuan (陶凯元) presented the document; Research Office director Zhou Jiahai (周加海), Civil Division No. 3 chief judge Li Jian (李剑) and Research Office deputy director Si Yanli (司艳丽) took questions; spokesperson Ji Zhongbiao (姬忠彪) moderated. The same day the Court’s official WeChat channel published a release note — drafting background, drafting considerations, and a part-by-part summary — followed by the full text. This brief translates the release note and adds DCC’s map of the 24 articles. DCC has published this instrument as three pages: the full English text is on the law page; the press-conference Q&A is translated separately; this brief is the Court’s own framing. For the instrument drawn as diagrams — the spine, the fault default, the notice flow, the product boundary and who answers for what — see the one-page visual reader.

What kind of instrument this is. A 法发-numbered “Opinions” is a judicial policy document (司法文件), not a judicial interpretation (司法解释, the 法释 series). The difference matters to anyone reading a Chinese judgment: a judicial interpretation is cited as the legal basis of a ruling; a 法发 document supports the court’s reasoning but the judgment cites the underlying statute. Inside the court system it is nonetheless binding guidance, and Article 22 tells higher courts to pull novel AI cases up for trial themselves — so this is the line trial courts will follow. The Court is candid that China has no dedicated AI law and that the Opinions therefore work entirely through existing statutes: the Civil Code, the Cybersecurity Law, the Data Security Law, the Copyright Law, the Anti-Unfair Competition Law (AUCL), the Consumer Protection Law, PIPL and the Civil Procedure Law, plus the Product Quality Law, the Road Traffic Safety Law and the Patent Law inside the articles.

Why it matters for a foreign company. Eight rules do most of the work. The default for AI torts is fault-based liability, not strict liability (Art. 3). “AI product” means a product with a physical carrier — a robot or a car, not a chatbot (Art. 9). The Civil Code Article 1195 notice-and-takedown rule now reaches generative-AI providers, including where a user prompted the infringing output (Art. 7). Training on lawfully public personal information within a reasonable scope is generally not an infringement, unless the individual expressly refused or the impact on them is major (Art. 6). Unconsented AI likenesses and cloned voices are personality-rights infringements, and the rule extends to the deceased (Art. 4). Algorithmic price discrimination is a tort, and AI-driven celebrity-impersonation selling is a punitive-damages fraud (Art. 10). A developer raising a non-infringement defense in an AI copyright case must disclose training-data sources, training records and the model’s operating mode (Art. 12). And anyone filing AI-generated pleadings or case-search reports must verify them and disclose the AI use (Art. 19). Two questions are deliberately left blank — the copyrightability of AI output, and whether unlicensed training on protected works is infringement — which the Court says it will revisit as consensus forms.

The translation below is DCC’s. Party-political framing is rendered as written; it is part of how the Court explains itself.

The Court’s release note

On 7 September 2026, the Supreme People’s Court held a press conference for the Opinions of the Supreme People’s Court on Trying Cases Involving Artificial Intelligence Disputes in Accordance with Law. SPC Vice President Tao Kaiyuan, Research Office Director Zhou Jiahai, Civil Division No. 3 Chief Judge Li Jian and Research Office Deputy Director Si Yanli attended and answered reporters’ questions. The press conference was moderated by SPC spokesperson Ji Zhongbiao. At the press conference, Vice President Tao Kaiyuan released the Opinions.

On 7 September the Supreme People’s Court publicly released the Opinions (hereinafter “the Opinions”). This is an important measure by which the Supreme People’s Court studies and implements Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era, fully implements Xi Jinping Thought on the Rule of Law, conscientiously implements General Secretary Xi Jinping’s important thinking on building China into a cyber power, fully implements the spirit of the 20th Party Congress and the plenary sessions of the 20th Central Committee, and serves and safeguards the accelerated development of new quality productive forces through high-quality judicial work. The drafting background, drafting principles and main contents of the Opinions are briefly introduced below.

I. Drafting background

At present, global innovation in artificial intelligence technology has entered an unprecedentedly active period. Intelligent technologies of ubiquitous connection and human–machine collaboration are compounding and releasing enormous energy, holding great opportunities while also facing governance challenges. In recent years the Party Central Committee with Comrade Xi Jinping at its core has attached great importance to the healthy development of artificial intelligence. In April 2025, at the 20th group study session of the Political Bureau of the 20th Central Committee, General Secretary Xi Jinping pointed out: “We must grasp the trends and laws of AI development, step up the formulation and improvement of relevant laws and regulations, policies and systems, application norms and ethical guidelines, build systems for technical monitoring, risk early-warning and emergency response, and ensure that AI is safe, reliable and controllable.” The Outline of the 15th Five-Year Plan again stresses improving laws and regulations, policies and systems, application norms and ethical guidelines in the AI field, and exploring the establishment of rules on the ownership of rights in AI-generated output and on the rights and responsibilities of developers, operators and users. The 15th Five-Year Plan for the Protection and Utilization of Intellectual Property recently issued by the State Council proposes improving the intellectual-property protection rules relating to algorithms, AI-generated output and the development of the platform economy, and implementing a program of mutual empowerment between intellectual property and AI.

How the judiciary should fully perform its functions, promote the healthy development of AI, serve national and social governance, and safeguard high-quality economic and social development is the question of the times facing the People’s Courts — a question that must be answered. To thoroughly implement the Party Central Committee’s series of major decisions and arrangements on AI development and governance, the Supreme People’s Court formulated these Opinions on the basis of a careful summary of adjudication experience, after extensively soliciting comments and conducting in-depth research and argumentation. The Opinions are problem-oriented and value-oriented; they focus on the prominent issues in AI-related adjudication that attract broad public attention, grasp the laws of judicial work amid rapid technological change, and support the innovative development of AI technology while drawing the legal “red lines” — so that the thousand-li horse of AI runs both fast and steady.

The Opinions are the first adjudication-rules document on AI issued by the nation’s highest judicial organ, and a contribution of “Chinese judicial wisdom” to the healthy and beneficial development of AI worldwide. In drafting, the Opinions received strong support and help from the relevant central bodies, representatives of the AI industry, experts and scholars, and other sectors of society; they are both a summary of adjudication experience and a distillation of the wisdom of all sectors.

II. Main considerations

First, resolutely implement the Party Central Committee’s decisions and arrangements, and serve and safeguard the in-depth implementation of the innovation-driven development strategy. By formulating an AI adjudication-rules document, the Opinions translate the spirit of General Secretary Xi Jinping’s important instructions on AI, and the Party Central Committee’s requirements on strengthening AI development and governance, into concrete practice. The Opinions strengthen judicial protection of AI innovation outcomes, innovating entities, innovative conduct and the innovation environment; delineate scientific and reasonable boundaries of conduct with clear rights and responsibilities; draw the legal “red lines” and bottom lines; promote technological exploration and industrial innovation through impartial and efficient adjudication; and release to the greatest extent AI’s enormous potential to empower high-quality economic and social development.

Second, strictly follow the spirit of the legislation, and ensure the uniform and correct application of law. AI is a new technology and AI-related legal issues are new issues; China has not yet enacted a dedicated AI law. Accordingly, following the principle of adapting to the times and making full and good use of existing legal provisions, the Opinions draw on the relevant provisions of the Civil Code, the Cybersecurity Law, the Copyright Law, the Personal Information Protection Law, the Civil Procedure Law and other laws to put forward conceptual and rule-based guidance for handling AI-related disputes in accordance with law — for example, clarifying the principles for attributing liability in AI-related tort cases, whether the use of already-public personal information in model training constitutes infringement, the definition of an AI product and the determination of defects, legal liability involving open-source software, and the patentability of AI-related inventions.

Third, coordinate development and security, and adhere to a people-centered approach and AI for good. The Opinions respect the laws of scientific and technological innovation and the development practice of the AI industry, and support scientific and technological innovation and industrial innovation through adjudication rules that accord with the spirit of law and policy and with the laws of technological development. Legal liability is determined accurately in accordance with law on the basis of the harm AI may cause in different application scenarios and the nature and magnitude of the risk. The Opinions respond properly to the rule conflicts, social risks and ethical challenges brought by the development of AI technology, attach importance to preventing and eliminating algorithmic discrimination, and prevent and regulate the abuse of AI technology that infringes the people’s lawful rights and interests. They give play to the guiding role of adjudication, actively steer AI toward enhancing the people’s well-being, and strengthen judicial protection of civil rights and interests in the digital era.

Fourth, remain problem-oriented and scenario-based, and enhance the document’s targeting and practical effect. Grounded in the adjudicative function, the Opinions focus precisely on the issues of broad public concern — “AI face-swapping and voice cloning”, “AI hallucination” torts, “online doxxing”, “big-data price discrimination”, autonomous driving, model training — and set out targeted substantive and procedural rules for AI-related disputes. Issues on which consensus is currently difficult to reach are deliberately left blank, to be clarified in an appropriate manner once further experience has accumulated and conditions are ripe.

III. Main contents

The Opinions consist of five parts and 24 articles. The main contents are:

First, they specifically set out the guiding ideology and basic principles for trying AI-related disputes. General Secretary Xi Jinping’s important statements on AI provide the fundamental guide for the People’s Courts in trying AI-related disputes and setting adjudication rules. Part I (Articles 1 and 2) sets out the guiding ideology to be upheld and proposes three basic principles — a people-centered approach, support for innovation and development, and a firm security line — ensuring that adjudication of AI-related disputes always keeps the correct political direction and effectively implements the requirements of equal emphasis on development and security, and of combining the promotion of innovation with governance in accordance with law.

Second, they regulate in accordance with law infringing conduct that uses AI. Part II (Articles 3–11) first clarifies the principles for attributing AI-related tort liability, and then addresses the determination and bearing of tort liability in different scenarios — the use of AI to infringe personality rights, personal-information rights and interests, and consumer rights and interests. It also clarifies the tort liability of generative-AI service providers, AI product liability, and compensation liability for traffic accidents involving autonomous vehicles and vehicles with assisted-driving functions.

Third, they provide for trying AI-related intellectual-property disputes in accordance with law. Part III (Articles 12–16) makes special provision for AI-related intellectual-property cases in five respects — determination of infringement liability, legal liability involving open-source software, patent grant and validity, performance of technology contracts, and norms for data use — actively responding to the new topics and challenges that rapid AI development brings to judicial IP protection, seeking a balance between encouraging AI innovation and safeguarding rights holders’ lawful interests, and providing clear adjudication rules for the high-quality development and high-level security of AI.

Fourth, they explore and improve procedural rules and punish unlawful and criminal conduct in accordance with law. Part IV (Articles 17–20) covers two matters. On the one hand, given the technical and specialized character of AI-related disputes, it explores and improves the rules for fact-finding and evidence review, and clarifies the regulation of conduct obstructing the judicial order, such as improper AI-assisted evidence-gathering and the submission of false materials. On the other hand, it lays down principles for punishing in accordance with law the various crimes committed using AI.

Fifth, they improve the mechanisms for trying AI-related disputes. The Opinions cannot cover every type of AI case; many questions require further exploration and accumulation in judicial practice, and AI governance is a systemic project that requires coordination among all parties and multiple measures. Part V (Articles 21–24) accordingly provides for the prevention and resolution of AI-related disputes, strengthened research and argumentation, case guidance and adjudication supervision, inter-departmental collaboration, and international judicial exchange and cooperation.

In his keynote speech at the 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance, General Secretary Xi Jinping stressed: “The faster AI technology develops, the more correctly its direction — upward, toward good, for the benefit of humanity — must be anchored, the more precisely the scale of regulation and governance must be grasped, and the more promptly the measures to prevent loss of control must be improved.” Going forward, the Supreme People’s Court will continue to implement Xi Jinping Thought on the Rule of Law, remain people-centered, take the issuance and implementation of the Opinions as an opportunity, persistently do a good job in trying AI-related disputes, give full play to the normative and guiding role of adjudication, promote AI that is upward, good and beneficial to humanity, and provide strong judicial services and safeguards for comprehensively advancing the building of a strong country and national rejuvenation through Chinese-style modernization.

DCC’s map of the 24 articles

The Court’s summary above is accurate but general. This is what each article actually does, with the statutory hook and the DCC coverage it connects to. Article numbers follow the SPC’s press usage; the full text is on the law page.

Part I — General requirements (Arts. 1–2)

Art. 1 is the guiding ideology. Art. 2 states the three principles, and buries three operative signals inside them: courts are to protect the individual’s right to know and right to choose in automated decision-making (the PIPL Article 24 vocabulary); to distinguish general-purpose from special-purpose and open-source from closed-source models when allocating liability among developers, providers and users; and to apply the criminal policy of combining leniency with severity — hard on intentional misuse, cautious on innovative conduct.

Part II — Torts (Arts. 3–11)

Art. 3 — Fault is the default. Unless a statute expressly imposes no-fault or presumed-fault liability, AI torts are decided under Civil Code Art. 1165(1). The fault factors are the ones that will matter in practice: application scenario, degree of autonomy, transparency, the preventive measures a developer or provider took and their technical feasibility, and the user’s capacity to foresee and control harm. Cheng Xiao’s analysis of which statutory duties actually ground fault is the closest reading of how those factors will be applied.
Art. 4 — Likeness, voice, reputation, the deceased. An unconsented AI-generated virtual digital likeness identifiable as a person, used or published, infringes the rights to name and likeness; a voice used as training corpus to clone timbre, intonation and pronunciation infringes the voice right (Civil Code Art. 1023); manipulating either to make untrue statements that lower social evaluation is defamation; and close relatives may sue over a deceased person’s likeness under Civil Code Art. 994. This is the civil-liability layer beneath the administrative rules in the Deep Synthesis Provisions and the AI Anthropomorphic Interaction Measures.
Art. 5 — Doxxing. Using AI to track and analyze a person’s public information (phone numbers, accounts, social media) in order to extract private information, or to leak it, or to disturb the tranquility of their private life, is a privacy tort; so is AI-enabled surveillance of private spaces. Note the mechanism: aggregation of public data with intent to pry is itself the infringement.
Art. 6 — Training on public personal information. Processing personal information the individual has self-disclosed or that is otherwise lawfully public, within a reasonable scope, for model training, is “generally not” an infringement if the individual has not expressly refused — but consent is required where the impact on the individual is major. The reasonable-scope test weighs purpose against model function, the type and sensitivity of the information, and the context and reasonably expected use at the time of disclosure. This tracks PIPL Art. 27 and Civil Code Art. 1036 and is the first time the top court has said so for training. Pair it with the training-data security standard GB/T 45652.
Art. 7 — Notice-and-takedown for generative AI. Where automatically generated content infringes personality rights and the provider fails to act promptly after notice (stop generating, block the prompt), it is liable for the harm. The notice must carry prima facie evidence and the rights holder’s real identity. Where a user maliciously induced the output, the user is liable, and the provider is jointly reachable under Civil Code Art. 1195 if it ignored the notice. This is the rule the Hangzhou Internet Court reasoned toward in China’s first AI-hallucination judgment.
Art. 8 — Injunctions. A personality-rights infringement injunction (Civil Code Art. 997) may issue not only against the actor but against a network service provider or generative-AI provider, ordering it to stop the service — proportionately.
Art. 9 — “AI product” means a physical thing. Product liability under the Product Quality Law applies to AI products with a physical carrier; the defect analysis weighs self-learning capability, updates, the degree of user control, and compliance with national and industry standards, with particular scrutiny of whether the producer gave truthful explanations and clear warnings of scenarios, limitations and foreseeable risks. Software services are outside. Again the Hangzhou court got there first.
Art. 10 — Consumers. Algorithmic unreasonable differential treatment on price or other terms for the same goods or services is a tort, judged by substantive harm to the rights to know, choose and deal fairly, by whether the terms were built from the consumer’s preferences, willingness and ability to pay or browsing history, and by good faith and commercial ethics. AI celebrity-impersonation selling that amounts to fraud draws punitive damages under Consumer Protection Law Art. 55. Read with the Platform Pricing Rules and the Algorithmic Recommendation Provisions.
Art. 11 — Cars. Autonomous and assisted-driving accidents are decided under the Civil Code and the Road Traffic Safety Law; defect claims go to producers and sellers; where a defect combines with driver fault, both are liable (Civil Code Art. 1172); overstated automation claims are actionable; and courts may order the data-controlling party to produce driving event records.

Part III — Intellectual property (Arts. 12–16)

Art. 12 — AI-generated content and copyright. Liability among developer, provider and user turns on service type, industry, training-data sources, each party’s participation, necessary measures and profits. A developer raising a non-infringement defense is ordered to produce training-data sources, training-process records, the model’s operating mode and the scientific basis. A user who knows of a prior work and generates a substantially similar one without a defense is liable. AI-enabled counterfeiting, false advertising and traffic or order inflation attract liability — the rule the courts built case by case in the AI seeding-post decision.
Art. 13 — Open source. Exemption is available to open-source developers and providers, weighed by license type, the limitations on rights, security and compliance measures, and disclosure; a contributor who supplies code modules free and publicly explains their function and security risks may be held not liable for a downstream user’s infringement.
Art. 14 — Patents. AI-related inventions are patentable subject matter when they use technical means following natural laws to solve a technical problem; a natural person who made a creative contribution to the substantive features is the inventor; sufficient disclosure is measured by the person of ordinary skill.
Art. 15 — Technology contracts. Breach is judged against the contract, the characteristics of AI R&D and whether the developer made reasonable efforts.
Art. 16 — Data. Lawfully obtained data rights are protected; datasets that qualify as compilation works go to copyright, trade-secret datasets to the AUCL, and other datasets to AUCL Art. 13 — the data clause the Beijing Internet Court applied for the first time in 2026. Algorithmic collusion and abuse of dominance, and data poisoning, malicious annotation and adversarial-sample attacks, all attract liability.

Part IV — Procedure and crime (Arts. 17–20)

Art. 17 gives courts the tools for opaque systems: investigation on application or on the court’s own motion, evidence preservation to fix technology, an adverse inference against a party that withholds documents or electronic data, and people’s assessors, forensic appraisers, expert assistants and technical investigation officers.
Art. 18 sets the review points for electronic evidence, big-data reports, blockchain-preserved evidence, and — importantly — for AI output offered as proof of infringement: prompt design and its influence, degree of similarity, consistency on repeated tests, and the model’s training, algorithm and filtering.
Art. 19 sanctions sham litigation built on manipulated AI output (including by deleting or altering the labels required under the AI content-labeling measures), routes AI-fabricated evidence to Civil Procedure Law Art. 114, and imposes the duty to verify and disclose AI-generated pleadings and case-search reports.
Art. 20 lists the offenses — fraud, insult, defamation, damaging commercial reputation, infringing citizens’ personal information, illegally obtaining computer-system data, obscene materials — and adds a specific one: defeating an assisted-driving system’s monitoring with aftermarket accessories.

Part V — Mechanisms (Arts. 21–24)

Diversified dispute resolution with industry regulators and specialized mediators (Art. 21); elevated jurisdiction so that novel, complex or rule-setting AI cases are tried by higher courts, plus the People’s Court Case Database (Art. 22); judicial recommendations and coordination with the cyberspace, public-security, procuratorial and market-regulation authorities (Art. 23); and cross-border AI and data cases handled with an eye to “rule alignment and mutual learning” (Art. 24).

What is not in it

The Court says so itself: the copyrightability of AI-generated content and the characterization of unlicensed training on others’ works were too contested to settle, and the Opinions make no provision on either. Both remain governed by the trial-court decisions that exist today, and both are flagged for future rules.

Source: 最高人民法院新闻局 (Supreme People’s Court Press Bureau), 《最高法发布〈关于依法审理涉人工智能纠纷案件的意见〉》, 最高人民法院 (Supreme People’s Court) WeChat Official Account, September 7, 2026. Original

— Not legal advice.