In February 2026 the Cyberspace Administration of China carried out a campaign against accounts that publish AI-generated synthetic information without adding an AI label and use false content to deceive and mislead the public, urging website platforms to investigate and rectify thoroughly. According to the report, 13,421 accounts were disposed of in accordance with law and agreement, over 543,000 pieces of unlawful information were cleared, and several typical cases were published.
Enforcement points: from “labelling” to “deceiving and misleading”
The campaign’s core focus was twofold. First, the absence of an AI label; second, the use of false content to deceive and mislead the public. According to public reports, some accounts used AI face-swapping and voice cloning to impersonate public figures such as athletes, actors, presenters and entrepreneurs and publish false statements.
This combination is worth noting: enforcement did not stop at the formal violation of “no label” but extended to the substantive consequence of “deceiving and misleading”. This means that even with an AI label, content that still causes misleading may be held to account; conversely, unlabelled content that causes misleading carries the highest risk.
Relationship to existing rules
The Measures for Labelling AI-Generated Synthetic Content already impose requirements for explicit and implicit labels. This campaign is another round of concentrated enforcement after those measures took effect. See enforcement of synthetic-content labelling and our note on the verification duty of generative AI users.
In the same period, the Cyberspace Administration of China launched the “Qinglang 2026 — Creating a Festive and Harmonious Spring Festival Online Environment” campaign, focusing on prominent online-ecosystem problems during the Spring Festival and the travel rush. AI-generated disinformation was listed as a key target.
Directions for platforms and companies
First, full coverage of labelling capability. Platforms should confirm that all features supporting generated synthetic content — images, video, audio and text — are equipped with labelling capability, avoiding a situation where “some are labelled, some are missed”.
Second, priority control of face and voice content. Content impersonating public figures mostly relies on face-swapping and voice cloning, so dedicated detection and interception should be set up. See our note on face-recognition filing and the sixteenth batch of deep-synthesis filings.
Third, retention of the account-disposal mechanism. Platforms should keep complete records from discovery to determination to disposal, for later verification. See reporting channels for AI misuse.
Fourth, risk alerts during holidays and peak periods. Rumours spike during the Spring Festival and travel rush, so platforms should deploy dedicated inspections in advance.
Pointers for content creators and MCNs
Accounts operating with AI-generated content should make “labelling” a fixed step in the publishing workflow; those imitating others with AI likenesses or voices need authorisation and must avoid misleading statements. See our note on the AI face-swap drama portrait-rights case and the regulation of digital-human streamers in live-streaming e-commerce.
The disposal of 13,421 accounts shows that the labelling duty has moved from “declaration of rules” to “routine enforcement”. For companies, labelling is no longer a debatable option but a technical precondition for publishing content.
This article is based on public information including the Cyberspace Administration of China’s report on the campaign against the spread of AI content lacking labels and false information. It is for general reference only and does not constitute legal advice.