China's Supreme People's Court issues its first national rules for judging AI disputes

On September 7, 2026, China's Supreme People's Court released its Opinions on Trying Cases Involving Artificial Intelligence Disputes in Accordance with Law (Fa Fa [2026] No. 10). The document has five parts and 24 articles and is the first set of AI adjudication rules issued by China's highest court. It rests on three stated principles: putting people first, supporting innovation and development, and holding a firm safety bottom line.

The opinions tell lower courts how to apply existing law to AI cases. Article 4 treats unconsented AI face-swapping and voice synthesis as infringement: a recognizable digital avatar built from someone's name or likeness violates their personality rights, and a synthetic voice mimicking a person's vocal characteristics infringes their voice rights. Article 7 applies the civil-law notice-and-takedown rule to generative AI, so a provider whose system automatically produces content that harms someone's reputation or privacy is liable if it fails to take necessary measures promptly after notice. Article 10 treats unreasonable algorithmic price discrimination based on personal data as a consumer-protection violation, and Article 11 places product liability on manufacturers for defects in autonomous and assisted-driving vehicles that cause accidents.

On training data, the court says processing personal information that is already lawfully public, within a reasonable scope, to train AI models generally is not infringement where the individual has not explicitly refused. Article 13 lets courts exempt open-source AI developers from liability where code is provided free with transparent safety disclosures, and Article 14 says AI-assisted inventions can be patented when there is human creative contribution. On copyright infringement, courts are told to weigh the type of AI service, the sources of training data and each party's degree of participation when allocating responsibility among developers, providers and users.

The opinions are judicial guidance under laws China already has, not the dedicated AI statute China has yet to pass, and they do not settle whether purely AI-generated works can be copyrighted. How much they change outcomes will depend on how lower courts apply open-ended standards such as "reasonable scope" and "timely" measures. Still, they give AI developers and platforms operating in China a single national reference for liability that until now had to be pieced together from scattered local rulings.