On November 6, 2026, the Cyberspace Administration of China (CAC, 国家互联网信息办公室) published the bulletin 网信部门从严查处利用AI技术假冒仿冒公众人物问题, reporting a concentrated clean-up of AI impersonation of well-known public figures on Chinese websites and apps. Cyberspace departments had required platforms to clear more than 47,000 unlawful items and to dispose of 3,961 accounts.
Operators of WeChat, Douyin, Xiaohongshu, Bilibili and Kuaishou serving users in China, providers of public generative-AI and deep-synthesis services in China, and brands and livestream sellers promoting goods on Chinese platforms with AI-generated likenesses of public figures must immediately review listings and accounts for unauthorized AI impersonation of public figures used for product promotion, health rumors, follower bait or vulgar content, remove the material, dispose of the accounts, and keep deep-synthesis labelling and content-management controls in place, or face CAC removal of the unlawful content and disposal of the accounts, and an investigation of website and app platforms where the problem is prominent. The November 6, 2026 bulletin names no fine amount and sets no later compliance date.
What does the November 6, 2026 CAC bulletin change?
The bulletin records a finished sweep, not a new instrument: platforms were pressed to clear unlawful information and accounts that used AI to impersonate public figures, and the four patterns below are the cases CAC chose to publish as a warning.
| Pattern | Platforms named | What the accounts did | Outcome in the bulletin |
|---|---|---|---|
| Unauthorized product promotion | WeChat, Douyin, Xiaohongshu | AI likeness or voice, or a family member's photo, used to sell goods | Accounts disposed of under law and platform rules |
| Health rumors | Stolen portraits plus AI effects claiming illness, including Alzheimer's | Accounts disposed of under law and platform rules | |
| Follower bait | Douyin, Xiaohongshu, Bilibili, Kuaishou | First-person views, life stories and emotional scripts in a public figure's voice | Accounts disposed of under law and platform rules |
| Vulgar content | Kuaishou | Synthetic scenes of a public figure in vulgar dance or intimate interaction, used to draw followers and sell | Accounts disposed of under law and platform rules |
The published totals are more than 47,000 items removed and 3,961 accounts disposed of. CAC states that platforms where the problem is prominent will be investigated, and that the clean-up of AI impersonation of public figures will continue.
Who is exposed to this CAC clean-up?
The exposed actors are the operators of WeChat, Douyin, Xiaohongshu, Bilibili and Kuaishou that serve users in China, providers of public generative-AI and deep-synthesis services in China, and brands and livestream sellers that promote goods on those platforms with AI-generated likenesses of public figures.
Deep synthesis, under China's Deep Synthesis Provisions, means AI-generated or synthetic text, audio, image and video, and it carries labelling, identity, content-management and algorithm-filing duties. The Provisions on the Management of Deep Synthesis in Internet Information Services are the standing text. The bulletin does not amend those duties. It shows CAC applying them to impersonation used for selling, rumors, follower growth and vulgar content.
Obsidian tracks 17 regulatory texts (in force, amended or in preparation) for the AI, Data and Digital Governance industry in China, as of November 2026. That count is the register of texts Obsidian holds for the industry in China. It is not a census of every measure in force.
What must platforms and sellers check immediately?
The required action is a review of listings and accounts, removal of the material, disposal of the accounts, and retention of deep-synthesis labelling and content-management controls.
- Search WeChat, Douyin, Xiaohongshu, Bilibili and Kuaishou listings and accounts for AI likenesses, cloned voices or family-member images of public figures used to promote goods.
- Remove posts that invent a public figure's illness, including Alzheimer's claims built from a stolen portrait and an AI effect.
- Remove first-person scripts that present a public figure's views, life story or emotional lines in order to draw followers.
- Remove synthetic vulgar dance or intimate scenes that use a public figure to attract followers or to sell.
- Confirm that public generative-AI and deep-synthesis services still label synthetic text, audio, image and video and still run content-management controls.
- Keep a record of what was removed and which accounts were disposed of, so a later CAC inquiry into a platform with a prominent problem has an answer.
Continuous per-jurisdiction monitoring is how a platform or a brand sees a CAC bulletin of this kind the moment it is published.
What happens if the material stays up?
The bulletin's stated consequence is removal of the unlawful content and disposal of the accounts, under law and under platform rules, plus an investigation where the problem is prominent. It states no fine, no named company and no appeal window. The four patterns above are the ones CAC published as the warning set.
Verify whether your accounts or your sellers' accounts on WeChat, Douyin, Xiaohongshu, Bilibili or Kuaishou match any of the four patterns, remove what matches, and brief trust-and-safety and the brand team on the November 6, 2026 totals. Obsidian keeps the CAC bulletin and the deep-synthesis duties in one jurisdiction view so the next platform notice is not read in isolation.
