China’s top court sets new rules for deepfakes, AI evidence and model training

China’s Supreme People’s Court has reportedly issued its first dedicated judicial guidance on AI disputes, covering deepfakes, synthetic voices, training data, algorithmic pricing and AI-generated evidence. The legal interpretation still needs to be checked against the court’s Chinese primary text.

Reports from ChinaTechNews and other outlets describe a 24-article document covering civil, intellectual-property and criminal matters. The reported guidance addresses unauthorised face-swapping and voice synthesis, digital recreations of deceased people, algorithmic price discrimination and liability in autonomous-driving incidents.

The reports also say people making substantial innovative contributions to AI-generated creations may be recognised as patent inventors, while publicly available personal data may be used for model training within reasonable limits without automatically constituting infringement. Parties relying on AI-generated legal materials or evidence would be expected to verify authenticity, with fabricated evidence potentially attracting penalties.

What is still unclear

The Supreme People’s Court document was not independently retrieved for this draft, and the available English-language reporting may simplify or translate important legal distinctions. The effective date, jurisdictional reach and treatment of foreign companies remain unconfirmed.

Our opinion

If the reported text is accurate, China is trying to turn several of the AI industry’s grey areas into courtroom rules rather than leaving them to platform policy. That is significant, but legal reporting is not the place for confident paraphrase: the Chinese primary document needs to be read before anyone treats these summaries as settled law.