Cyberspace authorities have reported the results of a recent concentrated crackdown on AI impersonation of public figures: in total, more than 47,000 pieces of illegal and non-compliant information were removed, and 3,961 accounts were dealt with. The four typical categories named in the notice almost all aim at "monetization": first, unauthorized use of AI to imitate celebrities' likenesses and voices, or even stealing photos of their family members to drive sales and traffic; second, stealing portraits and adding AI effects to fabricate rumors such as "illness" to mislead the public; third, impersonating celebrities in the first person to express opinions and using emotional language to lure followers; fourth, generating images of celebrities doing vulgar dances or engaging in intimate interactions to attract fans and make money. The cases involved multiple platforms, including WeChat, Douyin, Xiaohongshu, Bilibili, and Kuaishou.
A continuation of the two-month special campaignThis round of cleanup is not an isolated move; it is embedded in cyberspace authorities' continued governance of AI-related disorder in 2026. In September this year, the Cyberspace Administration of China launched the second phase of "Qinglang: Cracking Down on Disorder in AI Applications," making "impersonating or imitating others" a key focus. As of early September, more than 5.61 million pieces of content had been removed and more than 49,000 accounts had been handled. At the end of September, it rolled out nationwide a two-month "Qinglang: Cracking Down on Fake Personas in Short Videos" campaign, explicitly listing "using AI to impersonate people" as one of five key priorities and directly targeting fake endorsements and fake marketing using synthesized audio and video. The cases in this notice directly correspond to the targets of these actions.
Platform responsibility pushed to the forefront
