China's CAC drafts digital virtual human rules on labeling, consent, copyright

China's cyberspace regulator issued draft rules on April 3, 2026 to oversee the development of digital humans online, requiring clear labelling and banning services that could mislead children or fuel addiction. The draft, published by the Cyberspace Administration of China for public comment until May 6, 2026, would require prominent digital human labels on all virtual human content and prohibit digital humans from providing virtual intimate relationships to those under 18.

What it is

The draft measures are issued by the CAC and would govern internet information services delivered to the public in China through digital virtual humans, a category that spans virtual anchors, virtual idols, AI-driven customer-service avatars and motion-capture digital doubles. The rules would also ban using other people's personal information to create digital humans without consent and bar using virtual humans to bypass identity verification systems. The instrument functions as the image-layer counterpart to a separate five-ministry draft on anthropomorphic AI interaction services issued one week later, with the two rules together covering how virtual personas look and how they behave.

What it requires

Operators must obtain a natural person's independent consent before using sensitive personal information for modeling or image generation and must clearly explain the purpose and impact of that processing. Service providers, users and content-distribution platforms must display a conspicuous digital human label throughout the entire time the avatar is shown to users. The draft also bars infringement of intellectual property rights in the text, art, photography, music or audio-video works used to build a digital virtual human. Regulators can inspect algorithm filings and security assessments, issue warnings, order service suspensions and levy administrative fines of up to 200,000 RMB for non-compliance.

What security leaders should do

Security and compliance teams supporting any China-facing virtual-human, livestreaming or AI-avatar product should map which business units already use likeness-based avatars, since the draft reaches service providers, technology supporters, service users and distribution platforms across the commercial chain. Teams should begin tracking consent capture and withdrawal workflows for likeness and voice data now, build in persistent on-screen labeling by design, and confirm whether any avatar product has public-opinion or social-mobilization features that would trigger algorithm filing and security assessment before the rule is finalized.