Glossary · China
AI-generated content labeling measures (China)
Also known as: Measures for Labeling Artificial Intelligence-Generated Synthetic Content, AIGC labeling, GB 45438-2025
The Measures for Labeling Artificial Intelligence-Generated Synthetic Content are Chinese regulations in effect since September 1, 2025 that require AI-generated text, images, audio, video and virtual scenes to carry explicit labels visible to users and implicit labels in the file metadata, and that oblige distribution platforms to detect and label such content.
- AI regulation
- China
- Technical documentation
In one sentence
China’s labeling measures (since September 2025) require visible labels and metadata on AI-generated content, and platforms must detect it.
Example
An image generated by a Chinese AI service shows a visible “AI-generated” mark and carries the provider’s name and a content ID in its metadata; a social platform displays the label when the image is shared.
How it applies
- Explicit labels: Visible or audible labels — text at the beginning, end or in the middle of generated text, voice prompts in audio, prominent marks on images and in videos, and notices in virtual scenes.
- Implicit labels: Metadata in the file, including the content's attributes, the name or code of the provider and a content ID; digital watermarks are encouraged.
- Platforms: Services that distribute content must check for implicit labels, add prominent labels to content that is, may be or is suspected to be AI-generated, and let users declare AI-generated content. Removing or tampering with labels is prohibited.
- Standard: The mandatory national standard GB 45438-2025, also in effect since September 1, 2025, specifies the labeling methods.
- Technical documentation: Documentation that is generated with AI and published in China falls under the labeling duty. For AI products, the labeling mechanism must be described in the filing documentation and in the user terms.
Labeling measures vs. the EU AI Act
Both require machine-readable marking of synthetic content by providers. China adds mandatory visible labels for all generated content, duties for distribution platforms and a detailed technical standard — and applied them a year before the AI Act's transparency obligations. In the EU, disclosure by deployers is limited to deepfakes and certain public-interest texts.