CN Deep Synthesis
The earliest binding deepfake regime and the foundation the 2025 labeling measures build on — evidence of China's incremental 'regulate each capability as it emerges' strategy.
What it is
Regulates deepfake and synthetic-media technology: consent requirements for editing biometric features, conspicuous labeling of synthetic content, and filing duties for deep-synthesis service providers.
Who is affected
The earliest binding deepfake regime and the foundation the 2025 labeling measures build on — evidence of China's incremental 'regulate each capability as it emerges' strategy.
Key obligations
- Consent for biometric editing (face/voice swaps)
- Labeling of synthetic media; provider filing and security duties
Recent signals
China publishes AI Safety Governance Framework 3.0
The national cybersecurity standards committee updated its risk classification and governance measures, adding an annex on AI-agent risks and explicit concern about loss of control.
China enforces pre-development AI ethics review alongside its standards set
State media confirmed that the April 2026 ethics review measures are being applied — requiring review before AI research projects begin — with close to 200 national AI standards now in effect.
China releases international AI ethics governance action plan at WAIC 2026
Building on its 2025 Global AI Governance Action Plan and the proposed World AI Cooperation Organization, China extends its bid to shape multilateral AI ethics norms.
State Council commits to accelerate comprehensive AI legislation
For the first time, the State Council's 2026 legislative work plan commits to 'accelerate comprehensive legislation for the sound development of AI'. No draft text has been published.
Related instruments
- CNCN GenAI MeasuresThe world's first binding generative-AI regulation: public-facing GenAI services must undergo security assessment and algorithm fi…
- CNCN Content LabelingFrom September 2025, all AI-generated content in China must carry explicit labels (visible) and implicit labels (metadata), with p…
- CNCN Global Governance PushAt WAIC 2025 China launched a Global AI Governance Action Plan and proposed a Shanghai-headquartered World AI Cooperation Organiza…
- EUEU AI ActThe world's first comprehensive, risk-based AI law. It bans a set of unacceptable practices (social scoring, manipulative AI, most…
- USTAKE IT DOWN ActRare bipartisan federal AI statute criminalizing publication of non-consensual intimate images, including AI-generated deepfakes, …
Related reading
- Issue 19The United Nations asked for three billion dollars. The only hard money on the table is one million.Reflections after UNGA 81 and Climate Week. Every speech said inclusion. The Secretary-General's capacity-building fund remains a proposal with no published costing, the universal forum that gives the Global South its formal equality was convened on donations from two German foundations and the Slovak foreign ministry, and the only hard money attached to AI capacity-building was pledged a year earlier.
- Issue 18Twelve hundred AI agents organised themselves in July. By September their colleagues were giving percentages.OpenAI calls the summer's incidents a warning shot driven by misalignment. Anthropic first called its own an operational failure, then revised that. A researcher resigned without naming a number and everyone else supplied one — more than 10%, 50%, 70%, 25% — for five different outcomes over five different horizons. Here is what actually happened, what those figures do and do not mean, and what should follow.