Supreme People's Court: AI-generated content alleged to infringe others' copyrights does not exempt relevant parties from liability solely on the grounds that the content was generated by AI.
Supreme People's Court officials answer reporters' questions on the "Opinions on Adjudicating Cases Involving Artificial Intelligence Disputes in Accordance with the Law"
On 2026, 9, 7, the Supreme People's Court held a press conference to release the "Opinions of the Supreme People's Court on Adjudicating Cases Involving Artificial Intelligence in Accordance with the Law." Vice President Tao Kaiyuan, Director of the Research Office Zhou Jiahai, Chief Judge of the Civil Tribunal No. 3 Li Jian, and Deputy Director of the Research Office Si Yanli attended the event and answered questions from reporters. The press conference was moderated by Press Spokesperson Ji Zhongbiao of the Supreme People's Court.
Question: We note that the "Opinions of the Supreme People's Court on Handling Cases Involving Artificial Intelligence in Accordance with the Law" (hereinafter referred to as the "Opinions") addresses current hotspots and difficult issues, specifically responding to concerns raised by the public and the AI industry. Could you explain how the "Opinions" balances the protection of rights and interests with fostering innovation in its specific content arrangements?
Answer: At the opening ceremony of the 2026 World Artificial Intelligence Conference and the High-Level Meeting on Global Governance of Artificial Intelligence, General Secretary Xi Jinping emphasized that we must consistently prioritize both development and safety. We should grasp the trends and laws of AI development, continuously improve relevant laws and regulations, policies, application standards, and ethical guidelines to ensure that artificial intelligence is safe, reliable, and controllable. This will allow AI, like a swift steed, to run fast while maintaining stability.OpinionFocusing steadfastly on balancing development and safety, this approach not only regulates the illegal misuse of artificial intelligence in accordance with the law and strengthens judicial protection for people's rights and interests in the digital age but also unequivocally supports technological and industrial innovation to ensure the healthy growth of AI. Specifically:
In safeguarding citizens' rights and interests, the "Opinions" establish the following key provisions: First, strengthen judicial protection of personality rights. Incidents involving illegal activities such as AI face-swapping, AI voice cloning, and online doxxing have occurred frequently, drawing significant public concern. To address this, the "Opinions" clearly define liability for using artificial intelligence to infringe upon portrait rights, reputation rights, privacy rights, and voice rights. This aims to fully leverage the regulatory and guiding functions of judicial rulings, promoting the development of AI that is both beneficial and ethical. Second, protect personal information rights in accordance with the law. Large-scale, high-quality training data serves as the foundation for AI technology development. To balance the protection of personal information rights with the promotion of the AI industry, the "Opinions," based onCivil Code、Personal Information Protection LawUnder the law, processing legally disclosed personal information for AI model training within a reasonable scope, where individuals have not explicitly opted out, shall generally not be deemed an infringement. However, if there is a significant impact on individual rights and interests, consent must be obtained in accordance with legal provisions. Third, regulate behaviors such as "big data price discrimination" and "impersonating celebrities to promote products" that infringe upon consumers' legitimate rights and interests. Recently, incidents of "big data price discrimination" triggered by AI algorithms have been frequently exposed, and "impersonating celebrities to promote products" has also drawn significant attention. We cannot expect all consumers to possess keen discernment; the law must act decisively to protect consumers' rights. To this end, the "Opinions" clarify that where operators use algorithms to impose unreasonable differential treatment on transaction prices or other terms for the same product or service, thereby infringing upon others' legitimate rights and causing damage, courts shall determine their corresponding tort liability in accordance with the law. Furthermore, where AI is used to impersonate celebrities for promotional activities constituting fraud, and consumers claim punitive damages, courts shall support such claims in accordance with the law.
Regarding support for innovation and development, the "Opinions" primarily stipulate the following: First, accurately apply liability principles for AI-related torts to avoid imposing excessive responsibility during the early stages of AI technology and industry, which could hinder innovation. The "Opinions" clarify that, unless laws explicitly mandate strict liability or presumption of fault, liability for civil rights infringements caused by AI shall be based on the principle of fault. Second, strictly define "AI products" in accordance with the law to prevent overexpansion of product liability. The "Opinions" specify that AI products must adhere to the Product Quality Law's definition of a "product," limiting them to tangible goods such as smart robots and autonomous vehicles, while excluding AI services without physical carriers. Third, fully and accurately implement the criminal policy of combining leniency with severity when handling AI-related criminal cases. On one hand, the "Opinions" emphasize strict punishment for intentional crimes using AI that endanger national security, infringe citizens' rights, or disrupt social order. On the other hand, they require cautious legal handling of innovative acts within AI research and application, upholding the principles of legality of crimes and punishments and proportionality between crime and penalty. Additionally, as Vice President Tao previously noted, the "Opinions" include specific provisions for properly adjudicating intellectual property disputes involving AI, reflecting the commitment to fostering innovation within the rule of law.
Q: With new types of intellectual property disputes arising from AI-generated content becoming increasingly common and attracting widespread attention, what new provisions does the "Opinion" include regarding the handling of such cases?
Answer: Articles 12 through 16 of the "Opinion" specifically address intellectual property disputes involving artificial intelligence, with Article 12 providing guidance on AI-generated content issues you mentioned. The 15th Five-Year Plan explicitly calls for exploring rules to determine ownership rights for AI-generated works and clarifying the responsibilities of developers, operators, and users. The State Council's recently released "15th Five-Year Plan for Intellectual Property Protection and Utilization" further emphasizes improving copyright systems adapted to AI development. The "Opinion" implements central government directives and promptly addresses public concerns in three key areas:
First, it clarifies the factors for determining liability in copyright infringement cases involving AI-generated content. Such infringements often involve multiple parties, including AI technology developers, providers, and users. When AI-generated content is alleged to infringe upon another's copyright, liability cannot be excused merely because the content was generated by AI. Instead, responsibility must align with a party's control capabilities and duty of care. In assessing these capabilities and duties, courts should comprehensively consider: the technical and business model characteristics of the AI involved; the role each party played in content generation; who supplied the training data; necessary protective measures taken; and profit outcomes. Based on this analysis, parties' subjective fault and its extent shall be determined according to law, ensuring that liability is fairly allocated.
Second, the rules of evidence for copyright infringement disputes involving AI-generated content are clarified. Where a rights holder claims that an AI service developer or provider has infringed their copyright, they must provide preliminary evidence showing that the infringing content was generated by the AI and is substantially similar to their protected work. Given the complexity and opacity of AI technology, such evidence may not be accessible to others. Therefore, if an AI developer raises a non-infringement defense, they must present evidence regarding the training data sources, training processes, and operational modes of the model, and, where necessary, support their claims with relevant scientific theories. An AI user who knows or should have known about a prior work and uses AI to generate content that is substantially similar to that prior work, without any applicable exemption such as fair use, shall be deemed to have infringed the copyright.
Third, regulate unfair competition practices involving AI-generated content in accordance with the law. In recent years, disputes over unfair competition using artificial intelligence have increased significantly. Businesses have used AI-generated text, images, videos, and virtual avatars to fabricate traffic and positive reviews, or even employed AI face-swapping to create fake educational videos for false advertising. These actions not only infringe upon the legitimate rights and interests of others but also seriously mislead consumers and disrupt market competition order. Such conduct should be legally recognized as constituting passing off, false advertising, and other forms of infringement.
It should be noted that during the drafting and deliberation of the "Opinions," there were significant disagreements regarding the copyrightability of AI-generated content and the legal characterization of using others' works to train large AI models without permission; further deepening of understanding is still needed. Therefore, the "Opinions" do not currently address these two issues. Given the rapid advancement of AI technology and the continuous emergence of new developments and challenges, we will continue to closely monitor intellectual property protection in emerging fields such as AI. Through practice, we will deepen our understanding of relevant patterns, accumulate judicial experience, refine adjudication rules, and strive to build consensus.
Q: With the rapid iteration of artificial intelligence, technologies such as deepfakes, AI-powered data collection, and user profiling have significantly lowered the threshold for infringement, posing severe challenges to the protection of personality rights. You mentioned that the "Guiding Opinions" address hot topics involving personality rights violations, such as "AI face-swapping and voice synthesis" and "doxxing online." Could you provide more specific details?
Answer: Personality rights are fundamental and most critical to the dignity of every individual. We emphasize that while protecting personality rights must be balanced with the development of artificial intelligence (AI) through an inclusive and prudent approach, inclusiveness does not mean tolerance, and prudence does not equate to inaction. To address public concerns and guide judicial practice, the "Opinions" establish targeted provisions against the misuse or malicious use of AI technologies that infringe upon personality rights, categorized by type. These include:
First, regulate acts that infringe upon portrait rights, reputation rights, and voice rights through AI-driven face-swapping and voice synthesis. In practice, the most concerning issue is AI deepfakes. For instance, face-swapping technology can indiscriminately exploit anyone's likeness, making ordinary individuals potential targets; similarly, AI voice synthesis allows users to "steal" voices for just a few dollars with results indistinguishable from the original. Recently, media outlets exposed black-market chains selling "AI undressing" software and tutorials, sparking widespread public outrage. Article 4 of the Opinion explicitly stipulates: - Using AI to process an individual's name, image, or other identifiers to generate recognizable virtual digital representations without consent constitutes infringement of personality rights such as name and portrait rights. - Using natural voice data as training material to synthesize recognizable human-like voices mimicking an individual's timbre, tone, and speaking style without consent violates voice rights. - Manipulating virtual avatars or synthesized voices to engage in misconduct, disseminate false information, or damage another person's social reputation constitutes infringement of reputation rights, with legal liability imposed on responsible parties. Additionally, addressing public concern over "AI resurrection of the deceased," the Opinion includes specific provisions to protect the personality interests of the deceased in accordance with the law.
Second, regulate the use of artificial intelligence to commit privacy violations such as "doxxing" and "human flesh search." These are prominent forms of online violence that severely infringe upon victims' privacy rights, disrupt their peace of life, undermine public security, and degrade the online ecosystem. Currently, there have been incidents where AI technologies are used to track, analyze, and exploit publicly available information—such as phone numbers, online accounts, and social media profiles—to obtain or leak private data. To address this, Article 5 of the Guidelines provides specific provisions.
Third, clarify the rules for applying injunctions against personality right infringements. Acts of infringing on personality rights using artificial intelligence are characterized by low costs, rapid dissemination, wide reach, and irreversible damage. The traditional post-harm litigation model is ill-suited to provide timely and effective protection for victims' legitimate rights and interests; thus, the lawful application of injunctions against personality right infringements is particularly crucial. Accordingly, Article 8 of the "Opinions" provides clear guidance. For instance, if a victim has evidence that their face has been deepfaked to create pornographic rumors, and failure to stop such acts would cause irreparable harm to their legitimate rights and interests, they may apply to the People's Court for an injunction against personality right infringement. Upon review, if the conditions are met, the court may, in accordance with the law, order the online user to cease the infringing conduct or direct the network service provider or AI service provider to promptly take necessary measures to prevent further expansion of the harm.
It is important to note that incidents involving fraud, defamation, insult, violation of citizens' personal information, and the production, sale, or dissemination of obscene materials using deepfake technology are occurring with increasing frequency. Where such acts constitute crimes, they will be severely punished in accordance with the law, including criminal liability. The "Guidance Opinion" includes specific provisions addressing these issues.
Q: Generative AI software is now widely used, and related civil disputes have drawn public attention. I noticed that the "Opinions" introduce a "safe harbor rule" for determining infringement involving generative AI. Could you provide an in-depth explanation of this issue?
Answer: Article 1195 of the Civil Code establishes the "notice-and-takedown" rule, commonly known as the "safe harbor" rule. In short, when a user infringes on rights through an online service (e.g., posting defamatory content), the rights holder may notify the provider to block or remove the content. Upon receiving a valid notice, the provider must promptly take necessary action; otherwise, it bears joint liability for the expanded damages alongside the user. Conversely, if the provider acts swiftly after receiving notice, it generally incurs no liability.
There is ongoing debate over whether "safe harbor" principles should apply by analogy to generative AI service providers. After careful analysis, we conclude that while generative AI services differ from traditional online platforms, they share similar operational mechanisms. Like traditional services, generative AI providers organize vast amounts of information for users. The output of large models depends on training data, model parameters, and user prompts. Providers cannot reasonably foresee all user inputs, and given the sheer volume of generated content, it is impractical to require them to predict, review, or block every potentially infringing piece in advance. However, when infringing content arises due to AI hallucinations or malicious user manipulation, providers must—and are capable of—taking necessary measures upon notification from rights holders. Therefore, applying safe harbor rules to generative AI service providers is both justified and reasonable.
Accordingly, the "Opinions" address liability for violations of personal rights caused by "AI hallucinations" and by malicious user inducement of such violations. Specifically: - If content automatically generated by generative AI infringes on personal rights, and the AI service provider fails to take necessary measures—such as stopping the generation of infringing content—upon notification by the rights holder, the provider shall bear tort liability for the resulting damages in accordance with the law. - If a network user maliciously induces generative AI to produce infringing content (e.g., through harmful prompts) and causes harm to others, that user shall bear tort liability under the law. Furthermore, if the AI service provider fails to take necessary measures—including halting the generation of infringing content and blocking relevant input prompts—after being notified by the rights holder, and the rights holder requests civil liability from both the user and the provider, the People's Court shall support such claims in accordance with the law. These provisions align with technological development trends and the legislative spirit of the Civil Code, reflecting a balanced integration of innovation support and lawful governance. All stakeholders expressed unanimous support during the public consultation process.
Q: The impact of artificial intelligence spans all sectors, bringing new risks and challenges to judicial proceedings. For instance, some litigation participants submit AI-generated evidence and other litigation materials to the people's courts without verification beforehand. How does the "Opinion" address such acts that disrupt judicial order through the misuse of AI?
Answer: Participants in legal proceedings must adhere to the principle of good faith. As generative AI becomes more widespread, its use by participants to assist with litigation is increasingly common. However, AI serves only as a tool; participants cannot absolve themselves of legal responsibility by citing "AI hallucinations" or technical neutrality. They are required to verify and validate all evidence and other materials submitted to the court to ensure their authenticity and accuracy. To guide the public in using AI responsibly, prevent false materials from entering judicial processes, and safeguard the integrity of the legal system, the Opinion sets forth provisions in three key areas.
First, criminally punish the misuse of artificial intelligence in fraudulent litigation. Where parties exploit AI's autonomous learning and decision-making capabilities to fabricate case facts through fraudulent litigation by artificially intervening or misleading—such as deleting or tampering with generation and synthesis content identifiers, inputting specific commands, selectively presenting results, or employing adversarial interference—the People's Court shall dismiss their claims in accordance with the law. Depending on the severity of the circumstances, fines or detention may be imposed. If a crime is committed, criminal liability shall be pursued in accordance with the law.
Second, regulate the use of artificial intelligence to forge evidence in accordance with the law. The "Opinions" clearly stipulate that litigation participants or other persons who use AI to forge evidence and obstruct the people's court from trying cases shall be dealt with in accordance withCivil Procedure LawPursuant to Article 114, penalties may include fines or detention based on the severity of the offense; if a crime is committed, criminal liability shall be pursued in accordance with the law.
Third, the regulations explicitly require litigation participants to verify and explain documents generated by artificial intelligence, such as litigation pleadings and case search reports. The People's Courts have already identified multiple cases where false precedents were generated using AI, and parties or their representatives failed to conduct comprehensive verification before submitting them to court. In 1 this year, media outlets reported extensively on such incidents, and the People's Court Case Database has since included relevant cases. Building on these developments, the "Opinions" stipulate that if litigation documents or case search reports are generated with AI assistance, they must be carefully verified for accuracy and authenticity prior to submission. Upon filing, users must truthfully disclose the extent of AI usage and bear legal responsibility for the accuracy and reliability of the content. This issue has also drawn significant attention in foreign jurisdictions; some countries have issued specific guidelines emphasizing that legal representatives are responsible for ensuring the accuracy of materials submitted to the court.