On September 7, 2026, the Supreme People's Court issued the Opinions on the Lawful Trial of Artificial Intelligence-Related Dispute Cases (hereinafter referred to as the “Opinions”). The Opinions consist of five parts and twenty-four articles, responding to various specific legal issues arising in the practice of artificial intelligence applications. The rapid development of artificial intelligence technology is closely connected with the intellectual property system, which protects intellectual achievements. While regulating the trial of AI-related dispute cases, the Opinions also clarify various intellectual property issues arising from AI. This article focuses on the intellectual property-related provisions therein and, from the three dimensions of copyright, patents, and data protection, sorts out the relevant provisions and institutional logic of the Opinions.

I. Copyright

1.1 Allocation of Copyright Infringement Liability

Article 3 of the Opinions emphasizes that the principle of fault liability should generally apply to AI infringement liability. Article 12 provides that where AI-generated content infringes another’s copyright, infringement liability may involve three types of subjects: developers, providers, and users. The liability of these three types of subjects must be reasonably determined based on the type and characteristics of the AI service itself, the source of training data, the degree of participation of each party, the necessary measures taken, and the profits obtained. In light of the attribution principle in Article 3, the Supreme People's Court’s view is that in cases of AI copyright infringement, the allocation of liability among the parties should be comprehensively determined in light of the specific circumstances of the case, rather than directly attributing infringement liability to the party that directly uses the AI service or the party behind it that develops or provides the service.

1.2 Standards for Determining Users’ Infringement Liability

Article 12 of the Opinions clarifies that where a user knows or should know of the existence of a prior work, uses AI to generate a work substantially similar to the prior work, and has no reasonable defense, and the right holder requests that the user bear infringement liability, the people’s court shall support the request. Although this standard combines the user’s subjective fault with the objective result produced through AI, it still takes the traditional copyright law standard of “access plus substantial similarity” as its core, and there is no substantial changes in the standard for determining infringement liability.

1.3 Liability Exemption for Open-Source Software

In response to the fact that the artificial intelligence industry is highly dependent on the open-source ecosystem, Article 13 of the Opinions provides that in determining the infringement liability of open-source software developers, providers, and subsequent developers and providers, comprehensive consideration should be given to factors such as the type of open-source license agreement, the specific content of rights restrictions, security compliance measures, and the degree of information disclosure, and appropriate liability exemptions should be granted in accordance with law. In particular, where a subject provides certain code modules needed for artificial intelligence research and development in a free and open-source manner and publicly discloses the functions and security risks of the modules, when another person’s use of the module gives rise to an infringement dispute, the open-source software developer or provider may, based on the facts of the entire case, be found not to bear infringement liability. This helps maintain a balance between protecting innovation and safeguarding right holders, and avoids a chilling effect that suppresses the development of the open-source community.

1.4 Protection of Data Collections as Compilation Works

Article 16 of the Opinions extends copyright protection to the data field. Data and data collections that constitute compilation works or meet the constitutive requirements of other works shall be protected under the Copyright Law. This means that when the selection and arrangement of a data collection reflect originality, it may obtain copyright protection as a compilation work, regardless of whether the individual data therein are themselves protected, thereby providing a copyright-law protection path for creative investment in data elements.

II. Patent

2.1 Subject-Matter Eligibility of Technical Solutions

Article 14 of the Opinions aims to resolve disputes over whether artificial intelligence inventions and creations constitute patentable subject matter under the Patent Law. The standard it establishes is: where an artificial intelligence-related invention or creation employs technical means that follow natural laws, solves a technical problem, and achieves a technical effect conforming to natural laws, it shall be deemed subject matter protected by the Patent Law; except where it violates law, social morality, or harms the public interest. This formulation continues the traditional requirements of China’s Patent Law for protecting technical solutions, excludes pure rules of intellectual activity, mathematical methods, or business methods from patentable subject matter, and at the same time preserves patentable space for solutions combining artificial intelligence algorithms with hardware and processes.

2.2 Standard for Creative Contribution of Natural Person Inventors

AI participation in invention directly challenges the inventor system. Article 14 of the Opinions adopts the position centered on the creative contributions of natural persons: where a natural person uses AI to complete an invention or creation and makes a creative contribution to the substantive features of the invention or creation, that natural person shall be recognized as the inventor. In other words, AI is merely an auxiliary tool, and inventor status still belongs to the natural person who makes a creative contribution to the substantive features of the technical solution; if a natural person merely performs data input, parameter tuning, or mechanical operations and does not make a creative contribution to the substantive features of the invention, he or she cannot be recognized as the inventor. This standard both upholds the basic principle that only natural persons can be inventors and provides a basis for patent attribution rules in the AI era.

2.3 Authorization Conditions of Sufficient Disclosure

In response to the problem that AI patent specifications may use the algorithm black box as a reason to evade disclosure, Article 14 of the Opinions provides that where the description of the technical solution in an AI-related patent specification reaches the level at which a person of ordinary skill in the art can carry out the invention, it shall be deemed to satisfy the authorization conditions of sufficient disclosure. This means that the right holder may not refuse to disclose necessary technical content on the grounds that model parameters are trade secrets; the specification must reach the level at which a person of ordinary skill in the relevant field can implement the technical solution, otherwise it does not satisfy the authorization condition of sufficient disclosure under Article 26, Paragraph 3 of the Patent Law.

2.4 Prudent Determination of Liability for Breach of Technology Contracts

Article 15 of the Opinions addresses disputes involving AI-related technology contracts. In adjudicating disputes over technology development, transfer, licensing, consulting, and service contracts, liability for breach should be determined in accordance with the contract and with full consideration of the characteristics of artificial intelligence technology research and development, whether the technology developer has exercised reasonable efforts, and other relevant circumstances. Given that AI research and development is inherently exploratory and uncertain, the judiciary should adopt a prudent approach in such adjudication and should not simply treat failure in technology research and development as a breach of contract, thereby reflecting a policy orientation that encourages technological innovation.

III. Data Protection

3.1 Boundaries for Training Use of Publicly Available Personal Information

Compliance in training data is a core legal issue in the development of the artificial intelligence industry. Article 6 of the Opinions provides that, for artificial intelligence model training, processing within a reasonable scope personal information that an individual has disclosed themselves or that has otherwise been lawfully disclosed, and where the individual has not expressly objected, is generally not deemed to infringe personal information rights and interests; where it has a major impact on personal rights and interests, the individual’s consent should be obtained in accordance with law. This provision seeks a balance between protecting personal information rights and interests and reducing training compliance costs. In principle, publicly available information may be used for model training, but in circumstances with a major impact, individual consent is still required.

3.2 Tiered Protection System for Data Rights and Interests

Article 16 of the Opinions establishes a multi-tiered framework for protecting data rights and interests. First, where an AI developer lawfully obtains data through collection and generation, derivative creation, assignment, authorization/licensing, or other means and enjoys corresponding data rights and interests, the people’s court shall protect them. Second, data and data collections that constitute compilation works or meet the constitutive requirements of other works are protected under the Copyright Law. Third, data and data collections that constitute trade secrets are protected under the Anti-Unfair Competition Law. Finally, for data and data collections that do not constitute trade secrets, if the alleged infringing act violates Article 13 of the Anti-Unfair Competition Law, liability shall be borne in accordance with law. This article establishes a four-tier structure ranging from data rights and interests, copyright, to trade secrets and general clauses, so that different types of data can all find an appropriate basis for claims.

3.3 Data Monopoly and Adversarial Data Security

The Opinions also address the suppression of negative effects of data. Article 16 provides that where undertakings use data, algorithms, or other means to reach monopoly agreements or engage in abuse of market dominance, they shall bear corresponding liability in accordance with law; those who adopt technical means such as fabricating interference data, maliciously labeling data, or adversarial example attacks to harm the operational security of AI shall also bear liability in accordance with law. The former safeguards the market competition order for data elements, while the latter brings data quality and model operational reliability within the scope of judicial protection. This provision, together with Article 10’s regulation of big data price discrimination and Article 20’s provisions on data-related criminal liability, jointly constructs a judicial defense line in the field of data security.

3.4 Data Provision Obligations and Evidentiary Rules in Litigation

Article 17 of the Opinions clarifies that where a party in control of evidence such as documentary evidence or electronic data refuses to submit it without justified reason, and the opposing party asserts that the content of such evidence is unfavorable to the controlling party, the people’s court may find that assertion established. Article 18, addressing the particularities of AI evidence, requires that for evidence in the form of big data analysis reports, focus be placed on examining the source of the original data, cleaning rules, and the scientific nature of the analysis methods; for blockchain-preserved evidence, focus be placed on examining the authenticity of the data before it was uploaded to the chain and the reliability of the technical platform; where a party uses AI-generated content as evidence of infringement, comprehensive consideration should be given to factors such as prompt design, degree of similarity, consistency across repeated tests, and model training, algorithm design, and filtering mechanisms. Article 19 prohibits the use of AI to tamper with generated or synthesized content identifiers, engage in selective presentation of results, or conduct adversarial interference in order to obtain false evidence. These provisions alleviate, at the procedural level, the difficulties in fact-finding caused by the algorithm black box in the AI era.

In summary, against the background that dedicated legislation has not yet been promulgated, these Opinions provide clear judicial guidance for the trial of AI-related IP disputes and for industry compliance practice, and are worthy of study and research by IP practitioners.