Thesis
China is the only major AI power with a layered, sector-specific AI regulatory stack already in force. By mid-2026, three landmark rules — the Algorithmic Recommendations Provisions (March 2022), the Deep Synthesis Provisions (January 2023), and the Generative AI Interim Measures (August 2023) — are actively enforced. The 2024 AI Safety Governance Framework adds risk classification. For global builders, the practical implications concentrate on two areas: content labeling and compliance for any product accessible to Chinese users, and data localization for training data sourced in China. This page maps each regulation to its enforcement scope, key requirements, and the fastest English-language primary source.
Decision in 20 seconds
| Your concern | Relevant regulation | Best English source |
|---|---|---|
| Building a recommendation feed for China users | Algorithmic Recommendations Provisions (2022) | State Council English + China Law Translate |
| Generating synthetic media (faces, voices, video) | Deep Synthesis Provisions (2023) | State Council English; Xinhua summary same-day |
| Launching an LLM product in China | Generative AI Interim Measures (2023) | State Council English (full); China Law Translate (annotated) |
| Risk classification for your AI system | AI Safety Governance Framework (2024) | CAICT report (English summary); China Law Translate |
| Monitoring new AI rules as they appear | CAC official releases (Chinese-first) | Xinhua English (same-day summary); State Council English (48–72h full text) |
Current policy layers to track
| Policy layer | Why builders care | What to verify first | Best English-accessible path |
|---|---|---|---|
| Recommendation systems | Relevant when your AI product shapes feeds, ranking, or algorithmic recommendations for Chinese users. | Whether user control, disclosure, or recommendation logic becomes a deployment issue. | State Council English + China Law Translate |
| Synthetic media | Relevant when your product generates or transforms image, audio, video, identity, or avatar-like content. | Labeling, consent, and identity or voice usage expectations. | State Council English + Xinhua English summaries |
| Generative AI services | Relevant when you expose an LLM or gen-AI product to Chinese users or distribute AI-generated public content in China-facing contexts. | Whether the product is public-facing, who the users are, and whether content or security review enters the launch path. | State Council English + China Law Translate |
| Governance, standards, and data layer | Relevant when product risk classification, public-opinion impact, data handling, or standards alignment starts to matter. | Whether your deployment is moving from experimentation into regulated or higher-risk territory. | CAICT summaries, State Council English, China Law Translate |
English-accessible policy sources
| Source | URL | Coverage | Translation lag |
|---|---|---|---|
| The State Council English | english.www.gov.cn | Full text of major national AI regulations; State Council AI plans and white papers | 48–72h after Chinese release for major rules |
| Xinhua English | english.news.cn | Summaries of all major AI policy announcements; ministerial press conferences | Same-day for major announcements; summary-level only |
| CAC Official | cac.gov.cn | Primary text of all CAC rules (deep synthesis, generative AI, algorithm regs); enforcement actions | Chinese-only; State Council English covers major rules |
| China Law Translate | chinalawtranslate.com | Annotated translations of CAC, TC260, MIIT rules; technical AI standards; subscription for full archive | Days to weeks; unofficial but most technically accurate |
Compliance scope: when this becomes a real product constraint
The point of this page is not to replace counsel. It is to tell you when policy shifts from background context into launch risk. If your AI product is directly exposed to Chinese users, or your distribution path makes Chinese-user access central to the product, you should stop treating policy as passive reading and start treating it as a deployment review issue. If you are only monitoring China AI from outside that market, this page is mostly a signal layer: it helps you understand where the governance pressure sits and what kinds of updates are worth escalating internally.
FAQ
- Where can I track China AI policy updates in English?
- The State Council English (english.www.gov.cn) and Xinhua English (english.news.cn) publish official translations of major AI regulations, typically 48–72h after the Chinese-language release. China Law Translate provides annotated unofficial translations of technical standards. RadarAI's policy tracker maps each regulation to its source and notes translation lag.
- What are the key China AI policy layers builders should track?
- Track recommendation systems, synthetic media, generative AI services, and the broader governance or data layer. That is the practical builder map. You only need detailed legal review when your product crosses into direct China-facing deployment, public content distribution, or regulated data handling.
- What is the CAC and how does it affect AI development in China?
- CAC (Cyberspace Administration of China) is China's primary internet content regulator, with authority over AI-generated content, algorithm recommendations, and generative AI services. For AI builders, CAC rules require content labeling, security assessments before public launch, and data handling compliance for training data sourced in China.
- How do China's AI regulations compare to EU AI Act?
- The durable comparison is structural, not date-specific: the EU AI Act is a broad cross-market framework, while China AI governance is more layer-specific and content/deployment oriented. For builders, that means China policy often becomes concrete through deployment context and content handling, not just through one universal AI law lens.
- Is synthetic media regulated in China?
- Yes, since January 10, 2023. The Deep Synthesis Provisions require mandatory labeling for all AI-generated content (text, image, audio, video), consent for face/voice synthesis, and real-name registration for providers. Enforcement is by CAC. The regulation applies to any service accessible to users in China.
- Where do I find English translations of Chinese AI regulations?
- Official: State Council English (english.www.gov.cn), 48–72h lag. Summaries: Xinhua English, same-day. Technical/annotated: China Law Translate (chinalawtranslate.com). CAC primary text is Chinese-only — use State Council English for the official translated version of major CAC rules.
Companion pages in this cluster
| If your question is about… | Go to | What's there |
|---|---|---|
| How China AI compares to US AI overall | China AI vs US AI | Ecosystem comparison across models, compute, policy, and open source |
| Best English sources for China AI news | News Sources in English | Layer-by-layer routing guide for policy, model releases, funding |
| Weekly digest of China AI developments | Weekly China AI Digest | 15-minute Monday read covering models, policy, and funding signals |
| Chinese LLM API access from outside China | China AI API Access Guide | Qwen, DeepSeek, Kimi API international availability and signup |
| Broad China AI context and overview | China AI Overview | Topic definition, cluster routing matrix, start-here guide |
Quotable summary: The useful way to track China AI policy is by builder impact layer, not by memorizing a long rule timeline. Watch recommendation systems, synthetic media, generative AI services, and the broader governance/data layer; use State Council English for official text, China Law Translate for annotated technical detail, and Xinhua English for faster policy signals.