The CAC says platforms with prominent problems will be investigated and punished, and its sample cases ran on WeChat, Douyin, Xiaohongshu, Bilibili and Kuaishou.
China’s cyberspace authorities said they have removed more than 47,000 pieces of unlawful or rule-breaking content and acted against 3,961 accounts in a cleanup of AI-generated impersonation of public figures. The tallies come from a notice posted on November 6 on the website of the Cyberspace Administration of China (CAC), which Xinhua reposted the same day. Caixin Global carried the figures in its November 9 business brief.
The notice says authorities pressed websites and platforms to run concentrated cleanups and to deal strictly with unlawful content and accounts. It gives no campaign name, no start or end dates, no list of laws applied and no penalties. It does not say whether the numbers belong to the CAC’s September 30 campaign against fake personas on short video or to the broader AI-abuse campaign whose second phase the CAC reported in September.
For platforms and brands, the notice is a statement of what the CAC treats as a violation, with a warning attached: platforms with prominent problems will be investigated and punished through proper procedures.
The CAC lists four patterns of AI impersonation
The notice sorts its typical cases into four groups. The CAC notice names unauthorized use of a celebrity’s face or synthesized voice to sell products and draw traffic as the first of four patterns. That group also covers the stolen use of photos of a public figure’s relatives to promote goods. The second group takes a public figure’s photo and applies AI effects to invent rumors that the person has Alzheimer’s disease or another illness. The third group poses as a public figure and writes in the first person, offering opinions, personal stories or emotional appeals, to persuade users to follow the account. The fourth uses AI to place celebrities in vulgar dances or intimate scenes in order to gain followers and profit.
The sample accounts ran on WeChat, Douyin, Xiaohongshu, Bilibili and Kuaishou. They were handled under the law and under platform agreements, the notice says, without naming either. The CAC added that it will keep up the pressure, press platforms to take fuller responsibility and guard against AI misuse to protect users’ rights.
The notice sits on top of labeling rules and two earlier campaigns
The labeling framework dates to March 2025. The Measures for Labeling AI-Generated and Synthetic Content (《人工智能生成合成内容标识办法》) were issued by the CAC with the Ministry of Industry and Information Technology, the Ministry of Public Security and the National Radio and Television Administration, and took effect on September 1, 2025, according to the CAC’s question-and-answer page. Providers must add visible labels to AI-generated text, audio, images, video and virtual scenes, and hidden metadata labels to generated files. Users must declare AI generation and add a visible label before posting to the public, and no one may maliciously delete, alter or forge a label.
The CAC’s September 30 notice launched a two-month national campaign, 清朗·整治短视频虚假人设乱象, against fake personas on short-video platforms. AI-synthesized audio and video that impersonates public figures for false endorsement is one of its five targets. It tells platforms they may not rely on tags such as "fictional performance" or "AI-generated" to sidestep governance. EastFrontier described how that rule applied when WeChat Channels banned a seller that had cloned an actor’s voice. Earlier, a Xinhua repost of a CAC release on September 2 described the second phase of the 清朗·整治AI应用乱象 campaign, with 5.61 million items cleared, more than 49,000 accounts dealt with and about 2,400 websites and apps handled, and it named face-swapping and voice-swapping of public figures as a focus.
The totals are smaller than earlier tallies but count different things
The new numbers are modest next to earlier enforcement. A November 2025 CAC action punished more than 11,000 impersonation accounts, China Daily reported, after a deepfake of actress Wen Zhengrong’s face and voice appeared in merchants’ livestream sales. TechNode reported in February that regulators penalized 13,421 accounts and removed more than 543,000 pieces of content in a crackdown on unlabeled AI material. The CAC has not said whether the November 6 figures overlap with any of these totals, so they should not be added together or compared as a trend.
Three signals would show where enforcement goes next: a closing tally for the September 30 campaign, which the two-month length suggests would come around the end of November if counted from its announcement; any naming of platforms the CAC says have prominent problems; and a new case of the kind seen with Wu Yue, where a platform acts only after a public complaint.