Beijing's judges now have a rulebook for the deepfake era. On Monday, September 7, China's Supreme People's Court released the Opinions of the Supreme People's Court on the Trial of Cases Involving Artificial Intelligence Disputes — five parts, 24 articles — instructing courts nationwide on when an AI-generated face, a cloned voice, a hallucinating chatbot or a discriminatory pricing algorithm becomes somebody's legal liability. It is the first national judicial document on AI from China's highest court, and it lands in a country with no dedicated AI statute.
That absence is the point. Rather than wait for legislation, the SPC told lower courts to reach for tools already on the shelf: the Civil Code, the Cybersecurity Law, the Copyright Law, the Personal Information Protection Law and five other existing statutes. The rules were meant to “draw clear legal red lines” while supporting innovation, the court said, according to the South China Morning Post.
Faces, Voices and the Digitally Resurrected
The most consequential passage is Article IV, targeting what it calls “AI face-swapping and voice-mimicking” and “AI resurrecting the deceased.” Where a person's name or portrait is processed by AI without consent to generate an identifiable virtual digital image that is then used and published, the People's Court “shall support” claims for infringement of name and portrait rights. Voice gets identical treatment. If, without consent, “the voice of a natural person is used as training material to imitate the natural person's timbre, tone, and pronunciation style to generate a synthetic human voice that can identify that natural person,” the Opinions instruct, courts shall uphold a claim for infringement of voice rights.
Manipulating such a likeness to make someone appear to behave improperly or utter false statements is separately actionable as reputation infringement. And under Article 994 of the Civil Code, close relatives of a deceased person may sue over unauthorised digital resurrections — a live grievance in China, where AI recreations of the dead circulate widely.
Liability does not stop with whoever typed the prompt. Article VII makes generative AI providers liable in tort where they are notified — with preliminary evidence and the rights holder's real identity — that their system is producing infringing content and fail to take necessary measures such as ceasing generation. Users who “maliciously induce” a model into infringing output bear liability of their own, and Article 1195 of the Civil Code lets a plaintiff pursue both. Article VIII authorises pre-emptive injunctions: Xinhua's example is a victim of defamatory sexual deepfakes asking a court to intervene before the harm becomes irreparable.
Article X treats unjustified algorithmic differential pricing — “big data price discrimination” — as a tort, and authorises punitive damages under Article 55 of the Consumer Rights Protection Law where AI is used for “impersonating celebrities to sell goods.” “We cannot expect every consumer to become an expert at spotting deception,” said Zhou Jiahai, head of the SPC's research office, at Monday's press conference. “The law must step in promptly to protect consumers' legitimate rights and interests.” The document also covers autonomous-vehicle liability, AI training on personal information, doxxing, open-source exemptions, and a duty on litigants to verify and disclose AI-drafted filings.
Built on an Existing Scaffold
None of this arrives on blank ground. The Provisions on the Administration of Deep Synthesis of Internet-Based Information Services — 25 articles from the Cyberspace Administration of China and two ministries — took effect January 10, 2023. They were extended by the Measures for Labeling of AI-Generated and Synthetic Content, released by four agencies on March 14, 2025 and in force since September 1, 2025, requiring both visible labels and hidden metadata markers on AI-generated text, images, audio, video and virtual scenes. The courts had moved too: on April 23, 2024, the Beijing Internet Court decided China's first AI voice infringement case, awarding RMB 250,000 to a voice actor identified as Ms Yin after her recordings trained a text-to-speech product. The SPC has now nationalised that reasoning. Xinhua puts China's AI industry above 1.2 trillion yuan (about $177 billion) in 2025, up 40 percent year on year, across more than 6,600 companies as of June 2026.
Why It Matters
China, the EU and the United States are running three visibly different experiments. The EU AI Act is ex ante and administrative: its Article 50 transparency duties, requiring machine-readable marking of synthetic content and disclosure of deepfakes, apply from August 2, 2026, enforced by regulators with fines reaching 15 million euros or 3 percent of global turnover, rising to 35 million euros or 7 percent for prohibited practices. The United States has no federal AI statute at all — instead a patchwork of Tennessee's 2024 ELVIS Act, California's AB 1836 and AB 2602 covering digital replicas from January 1, 2025, and the federal TAKE IT DOWN Act signed in May 2025, whose platform obligations for intimate deepfakes bit on May 19, 2026.
China's move is a third path: neither a regulator's rulebook nor a legislature's statute, but instructions to judges on cases already arriving. It is retrospective rather than preventive, and works through private plaintiffs rather than agency enforcement — cheap to deploy and quick to update, but dependent on victims finding a courtroom. It also gives Chinese companies something the EU regime does not yet supply and the American patchwork actively frustrates: a single articulated answer to who pays when a model misbehaves.
The conspicuous silence is copyright. As the National Law Review noted, the Opinions never say whether AI-generated output is itself a protectable work — despite the Beijing Internet Court answering yes in Li v. Liu, and despite the SPC stripping Internet Courts of jurisdiction over online copyright disputes effective November 1, 2025. The court explicitly reserved contested questions for later, promising clearer rules once sufficient experience has accumulated. Watch for the first typical cases applying Article IV, for whether damages climb above the 250,000 yuan benchmark, and for whether a long-rumoured national AI law eventually codifies what the judges wrote first.
“We cannot expect every consumer to become an expert at spotting deception. The law must step in promptly to protect consumers' legitimate rights and interests.”— Zhou Jiahai, Head of the Research Office, Supreme People's Court