China CAC two-month short-video campaign targets fake personas, AI impersonation and label evasion

Key takeaways

• The campaign lasts two months. The release states no start date, end date or penalty amount.
• Problem area three covers AI-synthesised fake audio and video used to pose as public figures for false endorsement.
• Fake-persona videos without required content labels will be handled strictly, and platforms may not let labels such as AI-generated be used to evade governance.
• Fake "AI domineering CEO" and "wellness master" personas aimed at older people are singled out.

What did the CAC announce, and is it binding?

The CAC published a news release on 30 September 2026 saying that its office had recently issued a notice deploying a nationwide special action of two months, titled "Qinglang: Rectifying Fake-Persona Chaos in Short Videos". The release gives a stated purpose of promoting healthy and orderly development of the short-video industry. It names no start or end date.

It is a campaign notice, not a regulation. The release creates no new duty that we can find in its text; it describes problem areas and enforcement measures, and repeatedly says conduct will be handled "in accordance with law" (依法). We opened the release, not the underlying notice. A separate earlier campaign is covered in our article on the CAC Qinglang AI enforcement crackdown; this one concerns fake personas in short videos.

What does the campaign say about AI impersonation?

The campaign has five target problem areas. Four of them do not turn on AI: people posing as vulnerable groups, invented professional credentials, luxury and superstition personas, and staged conflict storylines. The third area is impersonation of figures using AI. Its first item reads: "通过AI合成虚假音视频等方式，假冒公众人物，实施虚假代言、营销等侵权违法行为。"

In English, that item targets AI-synthesised fake audio and video used to pose as public figures and carry out false endorsement, marketing and other infringing acts. The second item covers posing as parties or relatives in trending events, or reposting content to run "shadow accounts" (影子账号) that gather attention or charitable donations. The third item covers account names or profiles that link to a well-known public figure by adding or deleting words or symbols to suggest authorisation.

A fourth enforcement measure applies to specific groups. The release says videos that falsely build personas such as "AI domineering CEO" or "wellness master" to lure older people will be dealt with strictly: "虚假打造"AI霸总""养生大师"等人设诱骗老年人".

What do the label and platform measures require?

The release lists four enforcement measures. The first is tougher handling of violating accounts and platforms. It says the CAC will punish malicious marketing through fake personas, and mass production of uniform fake personas through MCN agencies, multi-account matrices or cross-platform reposting. Accounts of a serious nature will be banned.

The second concerns labelling at the point of publication. It says: "对于未按要求进行内容标注的虚假人设类短视频，依法从严处置" The same measure tells platforms to meet their content-review duties and says they may not use labels already applied, such as fictional dramatisation, AI-generated, marketing, repost or personal opinion, to evade governance requirements. The Chinese text reads: "不得借已标注虚构演绎、AI生成、营销、转载、个人观点等内容标签及无不良导向声明规避网络生态治理相关要求".

The third bars fake-persona videos from rankings and recommendation positions and says qualification checks will be strengthened. For fields that lack occupational classifications, qualification standards and certification requirements and that present serious scope for unlawful profit, it says they will be placed, by category, on the routine governance priority list. The release does not define the labelling standard itself, so the required label form depends on rules outside this release.

What does it mean for platforms, MCNs and brands?

For platforms, the release says local cyberspace authorities will supervise websites and platforms to check themselves against the stated problem areas and to examine key links and sections for gaps. For MCN agencies and creators, mass-produced personas distributed across several accounts or platforms are named as a target. In our reading, a brand that runs an AI-generated spokesperson, or a creator persona with a stated profession, faces more scrutiny on whether the persona is labelled and whether it implies a real person or a credential.

The release does not mention foreign companies or cross-border operators, and it names no penalty amount. We do not read it as changing what counts as lawful AI-persona use.

What should teams do now?

• List every persona, virtual presenter or AI-generated figure you publish on Chinese short-video platforms, with the account that runs it.
• Check that AI-generated content carries the label the platform requires, and do not treat a label as a substitute for accuracy about identity or credentials.
• Remove or correct personas that imply a real public figure's endorsement, or a doctor, lawyer or teacher title that the operator does not hold.
• If you commission content through an MCN, ask how many accounts and platforms each persona is distributed on, and who approves it.
• Watch for platform notices and any follow-up CAC announcement, since the campaign length is the only timing the release gives.

Which dates and time limits does the release give?

Only two timing points appear in the release. Everything else is unstated.

What we did not verify

What we opened: the CAC news release on cac.gov.cn, dated 30 September 2026, on 25 October 2026.

What we did not open: the notice that the release says the CAC issued, any platform implementation rules, any enforcement results, and any translation by another outlet. We cite the release by its section headings and numbering, and the English wording is our own rendering of the Chinese text.

What we refuse to claim: that the campaign creates new legal duties, that a start or end date or penalty scale has been set, that any particular company is under investigation, or what labels platforms will require beyond the release.

Related briefings

• AI Regulation News hub
• Asia-Pacific AI regulation tracker
• CAC Qinglang AI enforcement crackdown
• CAC draft anti-cyberviolence measures and AI labelling duties
• Beijing Internet Court on AI face swap and personality rights
• SAMR typical cases on AI and unfair competition