China's new rules, effective Jan. 1, require videos created using AI or VR tech to be clearly marked, make distribution of deepfakes a possible criminal offense
BEIJING/SHANGHAI (Reuters) - Chinese regulators have announced new rules governing video and audio content online …
Context & Ripple Effects
This rule is one link in a chain of Chinese content-control mandates that has been tightening since regulators required at least two human auditors to vet all audiovisual content back in 2017. Weeks after these labeling rules were announced, Beijing followed with separate rules requiring platforms to run AI-based moderation over everything posted online, effective that March.
What makes the Jan. 1 deadline significant is how it aged: the marking duty first imposed here became the template the Cyberspace Administration extended when it began regulating "deep synthesis" software in January 2023, and again when China mandated explicit-or-metadata labeling of all AI-generated material starting September 1, 2025. This 2019 rule is the origin point of China's synthetic-media control regime.
First-order effects
- Video and app operators serving Chinese users have until Jan. 1 to build visible marking into any AI- or VR-generated footage they distribute, adding a compliance pipeline on top of their existing editorial vetting.
- Distributors of face-swap and voice-clone content now face potential criminal liability if they circulate unlabeled deepfakes, shifting the legal risk from creation alone to distribution.
Second-order effects
- Platforms become the enforcement layer: with separate rules due in March requiring AI-based moderation, the same operators must simultaneously detect synthetic content and label it, pushing detection-tool vendors and app stores into the compliance chain.
- Developers of consumer video-editing and avatar apps must either embed labeling by default or gate features for the Chinese market, since their output is what distributors would be criminally liable for passing along.
Third-order effects
- If the cadence holds — 2019 video marking, 2023 deep-synthesis rules, 2025 metadata labeling — labeling obligations ratchet each time a new generation of generative tools ships, making provenance marking a permanent, expanding cost of operating any media platform in China.
- Criminal exposure for distribution points toward a structure where intermediaries, not just end users, bear legal responsibility for synthetic-media authenticity, a model other jurisdictions will watch as they draft their own deepfake rules.
The trend: China is constructing a layered synthetic-media regime in which every new class of AI content tool triggers a stricter labeling mandate, criminalizing distribution as the enforcement backstop.