China Mandates Labeling for All AI Generated Content

The Government of China implemented a new law on September 1, 2025, requiring all AI-generated text, images, audio, and video on social media to be clearly labeled. The regulations mandate a dual-layered approach consisting of explicit visible tags for users and implicit identifiers, such as digital watermarks embedded in metadata, to prevent misinformation, online fraud, and copyright infringement.

Led by the Cyberspace Administration of China, four government agencies drafted the rules as part of the 2025 Qinglang campaign to clean up the nation's cyberspace. Major platforms have already rolled out compliance features. WeChat and Douyin now require users to proactively label AI content and prohibit the removal of automated tags, while Weibo introduced a community reporting tool for unlabeled media.

AI developer DeepSeek is also implementing a mandatory labeling system that includes unique ID numbers and producer information in metadata to ensure content is traceable to its origin. The company has warned that altering or bypassing these labels could lead to severe legal consequences. To increase transparency, DeepSeek also released a technical guide regarding its model training and data usage processes.