CHINA Law and Practice Contributed by: Chuanhong Long, Huiqing Wang, Yingying Shen and Fan Li, CCPIT Patent and Trademark Law Office
7.3 Training Data, Model Development and Infringement Risk In China, the use of copyrighted works in training datasets gives rise to a real copyright infringement risk. Current regulatory rules already require providers of generative AI services to use data and base models with a lawful source in pre-training and optimisation training activities, and not to infringe the intellectual property rights of others. China currently has no clear general statutory TDM exception. It is therefore not safe to assume that “training” is automatically exempt. If a training data- set includes trade secrets, and those secrets entered the dataset through theft, breach of contract, unau- thorised access, or similar misconduct, liability may also arise under unfair competition law and, in serious cases, criminal law. As to remedies, a rights-holder may in principle seek cessation of infringement, damages, conduct preser- vation orders, and evidence preservation. Regulators may also require rectification by generative AI service providers. As for disclosure of complete training data- sets in litigation, China does not yet have a mature and uniform special disclosure regime; courts gener- ally still rely on case-specific evidence preservation measures, allocation of burden of proof, and court- directed evidence collection. 7.4 Enforcement Against AI-Enabled Infringement China already has legal mechanisms for pursuing infringement carried out through generative AI. Pub- lic judicial practice shows that courts have already applied existing copyright rules, personality rights rules, and related doctrines to AI-related disputes, including cases involving AI-generated images and AI-synthesised voices. Where AI is involved, courts generally still focus on traditional legal questions, such as whether the output is sufficiently similar to the protected subject matter, whether the platform had control over the generated result, whether complaint and takedown mechanisms were in place, and whether necessary measures were taken despite actual or constructive knowledge of risk. Platforms and tool providers may face second-
generally does not extend to the underlying factual data itself. As for scraping, crawling, and data extraction, Chinese law does not treat them as automatically unlawful in all cases. If a party, without authorisation, extracts plat- form data and creates a substantial substitute for the original platform’s service, harms the platform’s com- mercial interests, or disrupts competition order, the conduct may constitute unfair competition. By con- trast, where a party, with user authorisation, assists users within a reasonable scope in transferring data they are legally entitled to process, the conduct is not necessarily unfair competition. If personal informa- tion is involved, compliance with the Personal Infor- mation Protection Law and the Data Security Law is also required. 7.2 AI-Generated Works and Inventorship/ Authorship China currently takes a cautious, case-by-case approach to the copyrightability of AI-generated con- tent. If the final output reflects sufficient human intel- lectual input and personalised expression, a court may find that it constitutes a protected work. If, however, there is insufficient human creative contribution, the prospect of copyright protection remains uncertain. The author would generally still be the natural person, or other legally recognised rights-holder, who made the substantive creative contribution, rather than the AI itself. In the patent context, China’s current position is clearer. The Patent Examination Guidelines issued by CNIPA in 2025 and effective from 1 January 2026, expressly state that an inventor must be a natural per- son, and that an artificial intelligence name may not be listed as the inventor. Accordingly, AI cannot be named as an inventor in China. In practice, businesses typically designate the natural person who made the inventive contribution as the inventor, while allocat- ing application rights and ownership through employ- ment, commissioned development, or co-operative development agreements.
45 CHAMBERS.COM
Powered by FlippingBook