New Regulation on AI-generated Content

The Cyberspace Administration of China (CAC), together with three other departments, recently issued a new regulation on the identification of AI-generated content, which will take effect on September 1, 2025.

The rapid development of generative AI and deep synthesis technologies has made it easier to create synthetic text, images, audio and video. While these technologies boost economic growth and enrich online content, they also lead to risks, such as misinformation and damage to the online environment.

According to CAC, the regulation focuses on labeling AI-generated content to help users identify fake information and clarify service providers' responsibilities. It aims to regulate the entire content creation and distribution process, balancing costs with security. This will promote the safe use of AI in text generation, content creation, and design, while reducing the misuse of AI to spread false information.

Visible and hidden labels for AI content

"Labeling AI-generated content is now an international practice," said Jin Bo, deputy director of the Third Research Institute of the Ministry of Public Security. He said that the EU's AI Act, the U.S. REAL Political Advertisements Act, and the UK's AI Regulation Bill all require content labeling or watermarking.

In China, previous regulations have already addressed labeling, such as the deep synthesis provisions enacted in January 2023 and the interim measures on generative AI services in August 2023.

This time the new regulation introduces two forms of labels: visible and hidden. Visible labels use text, audio, or graphics to make users aware that the content was AI-generated, while the hidden labels are embedded in data files and not easily detectable.

Service providers offering AI-generated content must add explicit labels in certain cases and embed hidden labels in the metadata. Content distribution platforms are required to use technical measures to manage the spread of AI-generated content, according to the regulation.

"More than 300 generative AI products and 3,000 services have been registered in China. The regulation marks another step forward in building China's AI governance framework," said Zhang Zhen, senior engineer at the National Computer Network Emergency Response Technical Team/Coordination Center.

End-to-end governance

The regulation also requires application platforms to check whether services involve AI-generated content during approval reviews. Zhang said that the regulation covers multi-modal content such as text, images, audio, and video, and extend regulation from content creation to distribution.

Alongside this regulation, China will also implement a mandatory national standard — Cybersecurity Technology: Identification Methods for AI-Generated Content, effective on September 1. Fan Kefeng, deputy director of the China Electronics Standardization Institute, said this marks a new phase of combining technical standards with management rules in AI safety.

This dual approach clarifies responsibilities, improves oversight, and offers a complete framework for AI content governance in China, Fan added.

Source: Science and Technology Daily