China's top court sets first national rules for AI lawsuits
Without a dedicated AI law, China's Supreme People's Court tells judges how to rule on deepfakes, voice cloning, AI training data and self-driving crashes.
What happened: On September 7, 2026, China's Supreme People's Court issued a 24-article set of Opinions telling judges nationwide how to handle lawsuits involving artificial intelligence. It is the first national judicial guidance of its kind in China, and it arrives without a dedicated AI law: instead it tells courts how to apply existing statutes -- the Civil Code, Cybersecurity Law, Data Security Law, Copyright Law and Personal Information Protection Law, among others -- to disputes over deepfakes, voice cloning, digitally "resurrected" deceased people, AI training data, and self-driving car crashes.
Why it matters: For ordinary people, this hands real ammunition against a specific set of harms: someone who uses AI to swap your face, mimic your voice, or generate a synthetic version of a dead relative without permission can now be sued for violating name, portrait, voice or reputation rights. Doxxing built by AI-scraping someone's scattered public data counts as a privacy violation too. But the court sidestepped the industry's biggest open question -- whether an AI-generated image or video can be copyrighted at all -- leaving that fight to be settled case by case.
How it works, plainly: Judges are told to apply ordinary fault-based liability: unless a law sets strict liability, they must weigh how autonomous the AI was, how transparent its workings are, and whether developers took steps to prevent misuse. In copyright disputes, an AI company accused of training on stolen material must now produce its training data sources and process records to prove otherwise -- shifting the usual burden onto the developer. For self-driving cars, makers can be held liable for defects, split liability with distracted drivers, and be forced to hand over onboard data to establish what happened.
What's still unresolved: This is guidance for lower courts, not new legislation -- China still has no dedicated AI law. The gap is telling: a Beijing Internet Court had already ruled in 2025 that an AI-generated image was copyrightable, but the Supreme Court left that question out of its guidance and did not endorse it as a model ruling, favoring separate personality-rights cases instead. The top court also moved online copyright disputes out of the specialized Internet Courts that built this case law, a sign more clarity is still to come.