China Regulates AI-Generated Content: Towards a New Global Standard for Transparency?

‍An Objective: Regaining Control Over Information Flows

The approach is clear: regain control over an information environment that has become unstable by making AI generators, platforms, and Chinese citizens accountable. China thus establishes a standard for identifying synthetic content, which could well inspire other international regulations, starting with the European AI Act, which will not come into force until August 2026, and whose AI Office (European AI regulator) must develop standards for transparency of AI-generated content by the end of the year.

7 Key Measures to Remember

Here are the main provisions of this new Chinese regulation:

1. All types of content are concerned: text, image, audio, video, virtual scenes...

2. Mandatory double marking directly for AI generators:

• An explicit mention, visible to the user (warning panel, voice message, etc.)
• An implicit mention, invisible, in the form of metadata and digital watermarking.

3. Distribution platforms must read implicit mentions from AI generators and translate them into explicit mentions for end users.

4. In the absence of explicit and implicit marking, an explicit mention is imposed if the content is declared as synthetic by the user; here, the Chinese citizen is required to declare if they have, for example, used an AI generator located outside of China.

5. In the absence of any declaration or marking, platforms must themselves detect generation indicators and mark the content as "suspected to be synthetic".

6. The removal or falsification of mentions, both explicit and implicit, is strictly prohibited.

7. Tools allowing the removal of marks, metadata, and digital watermarks are also prohibited.

A Technological Bet: Digital Watermarking and Forensic Detection

To implement this identification system, China is betting on two key technologies:

• Digital watermarking, discreet and integrated at the source of the content. China thus joins the global scientific consensus around the use of watermarking, the most suitable technology for identifying AI-generated content
  
• Forensic detection, capable of spotting subtle indicators of AI generation.
  

These approaches pose real technical challenges: ensuring the robustness, resistance to falsification, and scalability of these tools is a major scientific challenge.

‍

‍

Innovation and Control: An Ethical Red Line

This regulation also reveals an assumed political vision: control prevails over innovation. By imposing strict obligations from the generation of content and throughout the content distribution chain, China forces technological actors to integrate compliance into their business models. A red line is thus drawn: innovation is desired, but never at the expense of trust and ethics.

Towards an International Standard?

While the world is still seeking its balance between innovation, freedom of expression, and digital security, China is taking a step ahead. Certainly, this regulation raises questions (cost, freedom, implementation), but it shows a firm will to define a standard in a field that is still largely unregulated.

It remains to be seen whether other countries will follow suit, and especially how. This ambitious regulation could, however, have been complemented by mandatory digital watermarking of trusted sites and media, in order to add even more certainty for Chinese citizens, as federal regulatory projects in the USA actually plan to do.

‍