China's top court sets first AI rules on deepfakes, pricing

On Monday, 7 September 2026, China's Supreme People's Court (SPC) released a 24-article judicial opinion on how courts should handle disputes involving artificial intelligence, according to state news agency Xinhua and reporting by the South China Morning Post. It is the first nationwide guidance of its kind, giving Chinese judges a common framework for lawsuits over deepfakes, cloned voices, algorithmic pricing and AI-generated misinformation — even though Beijing has yet to pass a dedicated law regulating artificial intelligence itself.

Rather than creating new criminal offences, the opinion tells courts how to apply four existing statutes — the Civil Code, the Cybersecurity Law, the Copyright Law and the Personal Information Protection Law — to disputes involving generative AI, the SCMP reported. The court described the guidelines as the first of their kind, intended to "draw clear legal red lines" while still supporting the growth of the domestic AI industry, in line with the government's current policy of encouraging AI development while managing its risks. The move comes as China races the United States for dominance in the technology, a context in which Beijing has generally preferred fast-moving administrative guidance over a single comprehensive AI law.

Deepfakes, cloned voices and 'bringing the dead back to life'

A central part of the opinion addresses AI-generated likenesses. According to Xinhua, the guidelines "make clear that people cannot use AI to create or distribute recognisable digital replicas of others without their consent, including cloned faces and voices." The SCMP reported that the court's examples extend to using AI to digitally "bring the dead back to life" — a service already sold commercially by some Chinese AI companies to grieving families — which the court said can infringe on legal protections covering a person's identity, image, voice and reputation, even when no defamatory content is involved.

The opinion also gives victims of sexualised deepfakes an explicit route to sue: it "offers legal recourse for someone whose face has been digitally altered and used to spread false and defamatory sexual claims," Xinhua reported. Under the new rules, an AI service provider can be found legally liable if it fails to "take timely action" once notified that its system has generated content infringing on someone's rights — a notice-and-takedown standard long used for user-generated content, now extended explicitly to outputs produced by generative AI systems themselves.

Algorithmic pricing, and a policy of 'leaving room' where there is no consensus

Beyond deepfakes, the guidelines also target "algorithmic price discrimination and AI-generated false information," Xinhua reported — language that points at the long-running Chinese consumer complaint known as "big data killing familiar customers" (大数据杀熟), in which platforms use a returning customer's own data to quietly charge them more than a new customer would pay for the same product. The opinion does not ban every algorithmic pricing practice outright. Instead, SPC Vice President Tao Kaiyuan told a press conference that the guidelines aim to "balance development and security" and mainly clarify procedural questions for judges hearing AI-related cases, rather than settle every substantive question in one pass.

• 24 articles, issued 7 September 2026 by the Supreme People's Court, applying the Civil Code, Cybersecurity Law, Copyright Law and Personal Information Protection Law to AI disputes
• Bans creating or distributing recognisable digital replicas — including cloned faces and voices — without the subject's consent
• Creates explicit legal recourse for victims of sexualised deepfakes and AI-generated defamation
• Holds AI service providers liable if they fail to act after being notified of infringing AI-generated content
• Targets algorithmic price discrimination and AI-generated false information, without banning dynamic pricing outright

For issues where consensus is currently difficult to reach, the guidelines leave room for clarification, allowing for further experience accumulation and appropriate clarification when conditions are ripe.

What it changes for AI companies serving Chinese users

President Xi Jinping had already flagged the direction of travel in July, urging officials at a national conference to "strengthen risk-awareness and ensure that AI is secure and controllable," language the SPC's opinion now translates into a concrete judicial standard rather than a political slogan. For any international company offering chatbots, voice-cloning tools or generative image and video products to users in China, the practical change is that a takedown request over an AI-generated deepfake or defamatory output now carries an explicit legal test: ignore a notice about infringing content, and a Chinese court can hold the service provider directly liable, regardless of where the underlying model was trained. The guidelines' language on algorithmic pricing is a second warning worth reading closely by any business running AI-driven dynamic pricing for Chinese customers, even one operating entirely from outside China — the same consumer-protection logic long used against "big data killing familiar customers" is now explicitly extended to AI-set prices.