On 23 June 2025 the National Technical Committee 260 on Cybersecurity (TC260) opened public consultation on six cybersecurity standard practice guides. All six address the same subject: methods for labelling AI-generated synthetic content.
From measures to guides: the implementation chain
The Measures for Labelling AI-Generated Synthetic Content create the obligation; the mandatory national standard GB 45438-2025 sets the method; and these six practice guides answer the question of how, specifically, to label. The chain narrows from the abstract to the concrete, and at each layer it reduces the room for a company’s own interpretation.
The six guides under consultation cover implicit metadata labelling for text, image, audio and video files, together with security protection and detection technologies. Labelling therefore has to be visible on the surface of content and machine-readable at the file level.
How explicit and implicit labels divide the work
The labelling regime distinguishes explicit from implicit labels. Explicit labels face users and tell a reader that content is generated. Implicit labels face systems, writing generation information into file metadata for platforms and regulators to verify.
The two solve different problems: explicit labels protect the right to know, implicit labels support provenance and accountability. See our notes on the Measures for Labelling AI-Generated Synthetic Content and on users’ verification duty for AI-generated content.
Links to other nodes in the standards system
Labelling methods are one element of the AI standards system and advance in parallel with other security and capability standards. See our notes on the basic security requirements for generative AI services and on the release of AI national standards.
Where companies can start
First, check whether the product’s labelling implementation covers both explicit and implicit forms; missing either leaves a compliance gap.
Second, watch the technical detail of metadata writing. Differences in file format affect the compatibility of implicit labels.
Third, bring labelling capability under version management, because the guides are still under consultation and the final text may change. See our note on enforcement practice for synthetic content labelling.
The success of a labelling regime turns on technical detail. Those who get the detail right will pass verification.
This article is based on the public consultation notice of the National Technical Committee 260 on Cybersecurity regarding six cybersecurity standard practice guides on labelling AI-generated synthetic content and public information and is for reference only; it does not constitute legal advice.