China has introduced a new regulatory framework governing the labeling of AI-generated and synthesized content. On 7 March 2025, the Cyberspace Administration of China (CAC), together with the Ministry of Industry and Information Technology (MIIT), Ministry of Public Security (MPS), and the National Radio and Television Administration (NRTA), jointly issued the Administrative Measures for the Labeling of AI-Generated and Synthesized Content (hereafter referred to as the Measures).
The Measures require network information service providers that generate or synthesize content using artificial intelligence technologies — including text, images, audio, video, and virtual scenes — to label such content in accordance with the new rules. The Measures apply to service providers falling within the scope of relevant existing regulations, including the Administrative Provisions on Algorithmic Recommendation of Internet Information Services, the Administrative Provisions on Deep Synthesis of Internet Information Services, and the Interim Administrative Measures for Generative AI Services.
The Measures specify two types of content labeling:
l Explicit labeling, which refers to marks presented in ways that are clearly perceivable by users, such as text, audio, or visual symbols displayed on the generated content or in interactive interfaces.
l Implicit labeling, which involves technical measures to embed identifiers within the file data of the generated content, not easily detectable by users.
Under the Measures, when a service provider’s content generation falls under specific circumstances outlined in Article 17 of the Administrative Provisions on Deep Synthesis of Internet Information Services, explicit labeling is mandatory. Additionally, providers are required to embed implicit labeling in the file metadata of AI-generated content in accordance with Article 16 of the same regulation. (The Article 16 and 17 can be found at the end of this article.)
To support the implementation of the Measures, China has also published the Compulsory National Standard — Cybersecurity technology - Labeling method for content generated by artificial intelligence, which will come into force on 1 September 2025, alongside the Measures.
Further guidance has been issued in the form of a practical guideline, the Cybersecurity Standard Practice Guide — Coding Rules for Implicit Labeling of AI-Generated and Synthesized Content by Service Providers, developed by TC260, China’s technical committee on cybersecurity standards. This guideline, released on 14 March 2025, provides coding instructions to help service providers implement implicit labeling in content file metadata.
In addition, TC260 is preparing a series of recommended standards and practice guidelines covering labeling specifications for various file formats and application scenarios. These supplementary documents will be released progressively following the publication of the Measures.
The introduction of the Measures and supporting standards imposes new compliance obligations on network information service providers, including overseas companies offering AI content services in China. Affected businesses are advised to closely follow regulatory developments and seek appropriate compliance solutions to adapt to the evolving regulatory environment.
Background Information:
Article 16 of the Administrative Provisions on Deep Synthesis of Internet Information Services
Deep synthesis service providers shall apply technical measures to add identifiers to information content generated or edited using their services, ensuring that such identifiers do not affect user experience. They shall also retain log information in accordance with laws, administrative regulations, and relevant national provisions.
Article 17 of the Administrative Provisions on Deep Synthesis of Internet Information Services
Where deep synthesis service providers offer the following deep synthesis services that may cause public confusion or misidentification, they shall place prominent identifiers in appropriate locations or areas of the generated or edited information content to inform the public that deep synthesis technology has been used:
l Text generation or editing services that simulate natural persons, such as intelligent dialogue and intelligent writing;
l Voice generation or editing services that synthesize or mimic voices, or significantly alter personal identity characteristics;
l Image or video generation or editing services that involve face generation, face swapping, face manipulation, posture manipulation, or other methods that significantly alter personal identity characteristics;
l Generation or editing services that create immersive, realistic scenes;
l Other services with functions that generate or significantly alter information content.
For deep synthesis services other than those specified in the preceding paragraph, service providers shall provide a prominent identification function and remind users of the deep synthesis services that they can enable such prominent identification.