China has moved to tackle one of the messiest parts of the AI boom: fake people. The country’s top internet regulator, the Cyberspace Administration of China, has published draft rules that would force companies to clearly label AI-generated “digital humans” and ban certain uses that could harm children, enable deception, or blur the line between real and artificial identities. The plan is to make it harder for AI personalities to pass as real people online.
The draft regulations, now open for public comment until May 6, require any service using human-like AI characters to display a prominent “digital human” label alongside the content at all times. That means virtual presenters, AI companions, influencers, customer service avatars, and any interactive AI persona must be clearly identified as synthetic. It’s a direct response to the explosion of AI deepfakes and virtual personalities that can easily be mistaken for real humans, especially as generative AI tools become more advanced and accessible.
David Adubiina
But the rules go further than just labeling. They would prohibit anyone from creating a digital human that resembles a real person without that person’s explicit consent, especially if it involves sensitive personal data. In other words, no more quietly training an AI model to look and sound like someone without their knowledge.
The draft also bans using digital humans to bypass identity verification systems, closing a loophole that could enable fraud or impersonation.
Another major focus of the legislation is protecting minors. Companies would be barred from offering “virtual intimate relationship” services involving digital humans to anyone under 18. Regulators are clearly worried about young users forming unhealthy attachments to AI personalities or being nudged into addictive, emotionally manipulative experiences. Providers are even encouraged to intervene and offer professional help if users show signs of self-harm or suicidal behaviour during interactions.
At the same time, the rules reflect China’s broader approach to AI governance. Digital humans would not be allowed to spread content that threatens national security, promotes social division, or violates what regulators call “socialist values.” Service providers are also urged to block sexually suggestive, violent, or discriminatory content generated through these virtual personas.
While this is still a draft, it looks like China is trying to set guardrails for AI identities before they spiral out of control. Compared to the United States, where regulation around AI deepfakes and AI companions is still fragmented and slow to take shape, this move shows a far more proactive stance. Whether one agrees with the scope of control or not, these proposals signal that at least one major government is taking the problem of AI fakes, digital deception, and emotional manipulation seriously and trying to legislate it before it becomes unmanageable.
David Adubiina
