Overview

• China’s law requiring explicit on-screen tags and hidden identifiers for AI text, images, audio and video took effect September 1 after being drafted by the CAC with three ministries.
• Major platforms including WeChat, Douyin, Weibo and Xiaohongshu have begun applying labels, with WeChat ordering users to self-declare AI content and banning removal or tampering of platform-applied tags.
• Douyin says it can use metadata to detect content origins, and Weibo has added a user option to report posts suspected of lacking required AI labels.
• AI firm DeepSeek will mark all content produced on its platform in China with visible notices and embedded metadata containing producer details and a unique ID, warning that attempts to bypass labels could carry legal consequences.
• Analysts caution that labels and watermarks can be stripped or evaded and note privacy and anonymity risks from metadata-based identifiers, even as regulators cite threats from deepfakes, fraud and copyright abuse.