China Tightens Rules on AI Digital Humans

China has moved to tighten oversight of AI "digital humans" as the market booms and ethical concerns mount. The Cyberspace Administration of China issued draft rules that demand clear labeling of AI-generated avatars and bar creation of a person's digital replica without consent. The sector was valued at 4.1 billion yuan in 2024, growing 85% year-on-year, and firms face fines from 10,000 yuan to 200,000 yuan for violations.

The draft rules focus on provenance, consent, and misuse mitigation for lifelike avatars often deployed across social media, e-commerce, and entertainment. Key operational requirements under consideration include:

• •explicit, visible labeling of any content produced by digital human systems
• •documented consent workflows for copying or recreating an individual's voice or likeness
• •mechanisms to prevent scams, misinformation, and socially destabilizing content
• •enforceable sanctions and reporting obligations for platforms and service providers

China's approach follows a familiar pattern: rapid commercialization followed by regulatory catch-up. The guidance targets the intersection of computer vision, speech synthesis, and generative models used to create emotionally persuasive avatars, sometimes described as "digital immortality." For practitioners, this raises engineering tradeoffs between realism and traceability: improving believability increases regulatory risk unless accompanied by robust consent management and metadata embedding for traceability.

Practical implications Product teams building or deploying avatar systems in China should prioritize consent capture, tamper-resistant provenance (watermarking or signed metadata), and moderation pipelines that detect impersonation and manipulative behaviors. Legal and compliance teams will need to scope data retention and user opt-in flows. Startups that monetize grief-support avatars must prepare for tighter scrutiny and possible limitations on certain use cases.

Expect consultation rounds and technical guidance that clarify labeling standards and acceptable consent artifacts. The enforcement emphasis will determine whether China sets a de facto global policy for avatar provenance and consent, influencing platform design choices worldwide.