Since September 2025, China’s largest social media platforms — WeChat, Douyin, Weibo and Xiaohongshu — have introduced mandatory labeling for content created with generative AI. Such materials are now accompanied by special labels and hidden watermarks in metadata.
The measure is being implemented under a new law passed to promote transparency and honesty in the use of neural networks. Users are required to mark AI content themselves, while platforms have tightened oversight: Douyin tracks file origins, WeChat prohibits removing labels, and Weibo allows users to report unmarked material.
The main goal is to combat disinformation and fake videos, which have become widespread as generative neural networks have spread. It is a step toward a new digital ethic in which honesty and trust in the source of information matter.
China has become the first country to introduce AI content labeling at the national level. In France and in some U.S. states, mandatory rules have already been introduced for certain areas, such as advertising and political materials. Google Pixel 10 smartphones support the C2PA technology, which automatically embeds provenance data in images, including whether AI was used in their creation or editing. All of this confirms that the world, albeit slowly, is moving toward a common standard of informational honesty.
