AI content labeling
Under Article 3 of the Measures for Labeling of AI-Generated Synthetic Content, any information such as text, images, audio, video and virtual scenes generated or synthesized using artificial intelligence technology counts as AI-generated synthetic content. There are only two kinds of label, explicit and implicit, and the two kinds are required of different parties at different stages — doing only one of them is not enough.
The article defines the content as follows: “AI-generated synthetic content means information such as text, images, audio, video and virtual scenes generated or synthesized using artificial intelligence technology.”
The same Article 3 splits labels into two kinds. An explicit label “means a label added in the generated or synthesized content or in an interactive scenario interface, presented as text, sound, graphics or similar, and capable of being clearly perceived by users”; an implicit label “means a label added to the data of the content file by technical means and not easily perceived by users”.
The scope of application sits in Article 2: service providers of internet information services that meet the situations set out in the Provisions on the Administration of Algorithmic Recommendation in Internet Information Services, the Provisions on the Administration of the Deep Synthesis of Internet Information Services and the Interim Measures for the Administration of Generative AI Services (referred to in these Measures as “service providers”) shall apply these Measures when carrying out labeling activities for AI-generated synthetic content.
Take stock of the outputs in your product: copy written by AI, generated illustrations, dubbed audio, digital human video and virtual scenes are all objects that need to be assessed.
Work out which outputs need both an explicit label and an implicit label, rather than only a corner mark on the display layer.
Put explicit means such as text or voice prompts and implicit means such as file metadata or digital watermarks on the requirements list separately.
Content created entirely by hand need not be forcibly labelled, but you should be able to explain its origin and how it was made if asked.
Thinking that only content that looks obviously fake counts, and overlooking that AI retouching, AI dubbing and AI-generated cover images are in scope too.
Reading labeling as “just add a watermark”, and missing the implicit label requirement.
Adding a notice only on the final display page, so that a file the user exports carries no label at all.
Reading “content” as the finished piece only, and overlooking that the interactive scenario interface itself can also carry a prominent notice.
Q: Does AI-assisted translation or AI polishing count as generated or synthesized content?
A: The article takes “generated or synthesized using artificial intelligence technology” as the core test. Whether a particular function falls inside depends on the actual shape of the service; for borderline cases, follow the practice of the local cyberspace administration office.
Q: Who are the labels for?
A: Explicit labels are for ordinary users and must be clearly perceivable; implicit labels are for situations such as platform verification and are written into the file data.
This entry belongs to the “China AI Compliance and Filing” cluster (42 entries in total): View all · Answers home
Who we are: a knowledge base and AI customer service team serving small and micro businesses, delivering implementations that take AI customer service through the national standard and the filing process.
Content on this site is compiled from publicly released regulatory texts for enterprises' own self-check reference; the specific filing requirements of the local cyberspace administration authorities prevail, and this is not legal advice.
This page is generated from the China AI compliance and filing MCP corpus — agents can call the same data directly.