China's cyberspace regulator issues draft rules for digital virtual human services

The Cyberspace Administration of China issued draft "Measures for the Management of Digital Virtual Human Information Services" on 3 April 2026, with a comment deadline of 6 May 2026.

As reported, the draft requires clear labelling of digital human use and explicit consent before a real person's likeness, voice or personal data is used to create a virtual representation. It prohibits using digital humans to evade facial recognition, voice recognition or other identity authentication mechanisms. It bans offering virtual intimate relationships to under-18s, including simulated family members or romantic partners, and services that promote excessive spending or harmful conduct. Content restrictions cover sexual innuendo, cruelty, horror and discriminatory material, as well as rumours, insults and defamation, alongside national-security provisions. Separate legal analysis of the same draft records algorithm filing and security assessment obligations for providers whose services carry public-opinion or social-mobilisation capacity, and fines up to RMB 200,000.

Why it matters for digital humans: China is the first major jurisdiction to regulate digital humans as a category in their own right, rather than folding them into general deepfake or synthetic-media rules. Two provisions travel well beyond China: the explicit anti-spoofing clause treats avatars as a biometric attack surface, which is a framing vendors selling into KYC-adjacent workflows should expect to see elsewhere; and the minors provision targets companion use cases directly. This is a draft, not law, and the comment period has closed without a published final text at the time of writing — treat it as a signal of direction, not a compliance obligation yet.