China enforces AI content labeling law across major social platforms

China has implemented a new law that requires social media platforms to label AI-generated content, which took effect on Monday. Major platforms activated their compliance features on the same day. Beijing's move is tied to the CAC's 2025 "Qinglang" campaign, which aims to combat misinformation and tighten AI oversight.

All AI text, image, audio, and video must carry explicit visible labels stating that it is AI-generated, while also carrying implicit identifiers such as digital watermarks or metadata. The law was drafted by the Cyberspace Administration of China in collaboration with the industry, public security, and broadcasting regulators. It was issued in March and has now taken effect.

The social media platform WeChat (also known as Weixin), with approximately 1.4 billion monthly active users, requires creators to declare AI-generated content, with the platform banning any tampering with labels. Another platform, called Douyin (known as TikTok in the Western hemisphere), with approximately 766.5 million monthly active users, requires visible labels on every AI-generated post and verifies sources through metadata. Weibo has added a new "unlabeled AI content" reporting option, with Xiaohongshu (also known as rednote) adding labels itself if users don't. Several services also remind users to judge unflagged content cautiously and to report it, while also reserving the right to remove mislabeled posts.

CAC flags penalties for using AI to spread misinformation or manipulate public opinion, with special scrutiny on paid online commentators. The law aims to combat misinformation, copyright infringement, and online fraud. Focus areas for this include deceptive marketing on short-video apps and providing stronger protection for minors. This encourages transparency online, although label evasion isn't entirely out of the question.

China is among the first major markets to legislate this mandatory AI-content labeling law, with similar moves emerging elsewhere around the world. The Internet Engineering Task Force has proposed an AI metadata header, with C2PA credentials shipping on devices like the Google Pixel 10. It is important to note that users can still sometimes circumvent safeguards.