China's Supreme People's Court, the country's highest judicial body, published a judicial interpretation on Monday telling lower courts how to handle civil cases that involve the misuse of AI tools. The document runs to 24 articles, according to the South China Morning Post, and it does not create a new statute. It tells judges which existing protections to reach for — a person's identity, their individual features, their reputation — when someone's face or voice shows up in material they never made.
### ⚖️ What the 24 articles cover
The interpretation groups together disputes over AI-generated video and audio, sexualized content, financial violations, digital privacy, and pricing based on surveillance. Deepfakes sit at the centre of it: realistic images, video and audio modelled on specific people. The court notes that the technology can now produce material that looks like "the dead returning to life."
As generative models get cheaper to reach, deepfakes are spreading quickly online, and fraud schemes, disinformation and abusive content are spreading with them. Si Yanli, deputy head of the Supreme People's Court's research office, said many people are seriously worried about deepfakes in practice. Face-swap technology makes it possible to collect anyone's images indiscriminately, so an ordinary citizen can become a victim; voice cloning makes it cheap to steal a person's voice and produce a convincing fake.
Courts across the country are instructed to weigh such disputes against the need to protect a person's identity, their individual traits and their reputation.
### 🔒 Doxxing in the same document
Beyond AI misuse, the interpretation also covers doxxing — obtaining other people's personal information and spreading it online. Si Yanli called it a conspicuous form of cyberbullying that seriously violates victims' privacy, disrupts their daily lives and undermines security in cyberspace.
Putting doxxing in the same document as deepfakes is not an accident of drafting. Both are cases where information about a person is taken and used without them, and grouping them tells judges to treat the harm, not the tool, as the thing being adjudicated.
### 🏛️ Why a court memo binds where precedent doesn't
China is a civil law jurisdiction, so its courts do not operate on binding precedent. Unlike in the United States, a lower Chinese court is not automatically bound by what another court decided earlier. Interpretations from the Supreme People's Court are the exception in effect if not in form: lower courts are obliged to consult them when preparing their own rulings. That is what turns 24 articles into a real barrier rather than a statement of concern.
The most informative thing in the document is the list itself. Surveillance-based pricing has nothing technically in common with a face swap. What they share is that a system used a person's data to produce something the person never agreed to — and filed together under that description, they become a dignity and consumer-protection problem rather than a technology problem. That is the move that lets existing civil law do the work without waiting for a legislature.
What the account of the document does not settle is where liability lands. A deepfake case has at least three candidates: the person who made the fake, the platform that carried it, and the service that made producing it trivial in the first place. Guidance aimed at protecting identity, traits and reputation tells a judge what was violated. It does not, on this description, tell them whose balance sheet pays — and that is the question that determines whether any of this changes how the tools are built.
A judicial interpretation is faster than a statute and easier to revise, which is the advantage here and also the thing worth watching. China has just shown how much ground its civil courts can be redirected to cover in a single Monday filing. The same instrument that now reaches a face-swap app reaches whatever the court decides to point it at next.