---
id: GOV-CN-001
title: "China's AI Regulation"
alternative_names:
  - "PRC AI Governance"
  - "Chinese Generative AI Measures"
category: "Safety & Governance → Regulation"
volume: VIII
chapter: 48
entry_type: "Standard Article"
level: 3
status: "Draft"
last_reviewed: 2026-10-18
version: "0.2.0"
prerequisites:
  - "GOV-LAW-001"
related_entries:
  - "GOV-EU-001"
  - "GOV-US-001"
  - "GOV-INTL-001"
  - "GOV-EXPORT-001"
---

# China's AI Regulation

*Legal status as of 2026-10-18. Instruments are described from the Chinese official text where it could be retrieved (CAC pages for the algorithm, deep synthesis and generative AI instruments), and otherwise from official-media reproductions (Xinhua) as marked. Translations of titles are unofficial.*

## 1. Concise Definition
China's AI regulation is a set of binding administrative measures, national standards and statute amendments, issued mainly by the Cyberspace Administration of China (CAC) with other ministries, that regulate specific AI applications (recommendation algorithms, deep synthesis, generative AI services, and labelling of AI-generated content). As of 2026-10-18 China has no single comprehensive AI statute in the sources reviewed; comprehensive AI legislation appears in official 2026 legislative plans but no draft was found.

## 2. Overview & Significance
China regulates AI through a layered, application-specific approach and has acted earlier than other jurisdictions on generative AI services. Its regime combines content-control obligations tied to state policy with technical-governance measures (algorithm filing, security assessments, content labelling). The approach contrasts with the EU's horizontal risk-tier statute ([[GOV-EU-001]](../vol08_safety_governance/GOV-EU-001_the_eu_ai_act.md)) and the US reliance on executive action ([[GOV-US-001]](../vol08_safety_governance/GOV-US-001_united_states_ai_policy.md)). Whether the regime is "stricter" or "lighter" depends on the dimension compared; it is more restrictive on content and more permissive of state-directed industrial promotion.

## 3. Historical Development
- **Algorithm recommendation provisions.** CAC Order No. 9 (jointly with MIIT, Ministry of Public Security and the State Administration for Market Regulation), dated 31 December 2021 and effective 1 March 2022 (CAC, 2021). Article 24 requires providers of services with "public opinion attributes or social mobilisation capacity" to file with the algorithm filing system within ten working days of starting service.
- **Deep synthesis provisions.** Joint order of the CAC, the Ministry of Industry and Information Technology and the Ministry of Public Security, issued 25 November 2022 (Order No. 12) and effective 10 January 2023, as published on the CAC site (CAC, 2022). Article 9 requires real-identity verification of users of deep synthesis services and management of deep synthesis content.
- **Interim Measures for the Management of Generative AI Services.** CAC Order No. 15, issued 10 July 2023 by the CAC jointly with the NDRC, Ministry of Education, Ministry of Science and Technology, MIIT, Ministry of Public Security and the National Radio and Television Administration; effective 15 August 2023 (CAC, 2023).
- **Measures for Labelling AI-Generated and Synthetic Content.** Jointly issued by the CAC, MIIT, Ministry of Public Security and the National Radio and Television Administration, effective 1 September 2025. A CAC statement reproduced by Xinhua on 14 March 2025 says a mandatory national standard, *Cybersecurity technology: Method for labelling AI-generated and synthetic content*, approved by the State Administration for Market Regulation and the Standardization Administration (2025 No. 3), takes effect on the same day (Xinhua, 2025). The standard number GB 45438-2025 comes from secondary law-firm summaries and was not seen in an official page [VERIFICATION NEEDED: standard number and CAC order number on the CAC original].
- **Cybersecurity Law amendment.** Decision adopted by the NPC Standing Committee on 28 October 2025 (18th session of the 14th Standing Committee), effective 1 January 2026 (Xinhua, 2025b). The text of the Decision, as reproduced by a security-industry site citing the NPC website, adds a new Article 20 under which the state supports basic AI research and key technologies such as algorithms, training-data and compute infrastructure, improves AI ethics norms, and strengthens risk monitoring and safety supervision; Xinhua describes the revision as focused on strengthening legal liability and raising penalties for some violations [VERIFICATION NEEDED: the npc.gov.cn page itself was not opened; article wording taken from a reproduction].
- **Legislative planning.** Xinhua reported on 11 May 2026 that the State Council General Office issued the *State Council 2026 Legislative Work Plan*, which "deploys" faster work on comprehensive legislation for the healthy development of AI (Xinhua, 2026). Media reports of the NPC Standing Committee's 2026 plan, published the same day, say it lists AI-related legislation as a preliminary-review item, to be researched and drafted "by the relevant parties" and scheduled "as appropriate" [VERIFICATION NEEDED: NPC plan text not read first-hand]. The Standing Committee work report to the NPC (March 2026, reproduced by Xinhua) says the committee will "strengthen research on legislation in fields such as artificial intelligence". No draft AI law was found in sources reviewed.

## 4. Technical / Conceptual Explanation
**Generative AI Interim Measures (Chinese text read at CAC).**
- *Scope (Article 2)*: services using generative AI to provide text, image, audio or video content to the public within the PRC; organisations that develop or apply the technology without offering the service to the domestic public are outside scope.
- *Content (Article 4)*: providers must comply with law and social ethics, must not generate content prohibited by law, and must take measures against discrimination, respect intellectual property and personal rights, and improve transparency and reliability.
- *Training data (Article 7)*: data and base models from lawful sources, respect for intellectual property, consent or another legal basis for personal information, and efforts to improve data quality.
- *Filing and security assessment (Article 17)*: services with "public opinion attributes or social mobilisation capacity" must undergo a security assessment and complete algorithm filing.
- *Effective date*: 15 August 2023.

**Labelling.** The 2025 measures distinguish explicit labels perceivable by users from implicit labels embedded in file data (such as metadata). A companion mandatory national standard sets implementation methods. According to Xinhua's account of the Measures, explicit labels are text, sound or graphics perceivable by users, and implicit labels (for example in file metadata) carry attributes such as the content's generated status and the provider's name or code.

**Compute thresholds.** The sources reviewed contain no FLOP-based regulatory threshold in these measures.

## 5. Types, Components & Classifications
| Layer | Instrument | Status as of 2026-10-18 |
| :--- | :--- | :--- |
| Algorithm governance | Algorithm recommendation provisions | In force since 2022-03-01 |
| Synthetic media | Deep synthesis provisions | In force since 2023-01-10 |
| Generative AI services | Interim Measures | In force since 2023-08-15; "interim" in title |
| Content labelling | Labelling Measures + mandatory national standard | In force since 2025-09-01 |
| Statute | Cybersecurity Law (amended) | In force since 2026-01-01 (amendment decision) |
| General AI law | none | Not enacted; in 2026 State Council plan and NPC research/preliminary items |

Related law not covered in depth: the Data Security Law and the Personal Information Protection Law, which the Interim Measures expressly require providers to observe.

## 6. Applications & Concrete Examples
Providers of public-facing generative models operating in mainland China must comply with the filing and labelling requirements; the CAC publishes lists of registered services, whose size is a fast-changing figure and is not reproduced here [VERIFICATION NEEDED: no count obtained from a CAC page]. Article 20 of the Interim Measures provides that for generative AI services originating outside the PRC and offered domestically in breach of the rules, the national cyberspace authority shall notify relevant bodies to take technical and other necessary measures; whether and how this has been applied was not checked.

## 7. Advantages, Limitations & Unresolved Questions
- *Reported strengths*: early, concrete obligations (labelling, filing), supervisory capacity.
- *Limitations and criticisms*: content obligations are tied to state-policy requirements, which are criticised as constraining expression; regulatory scope relies on administrative discretion; enforcement data are not systematically public in sources reviewed.
- *Unresolved*: whether a comprehensive AI law will be adopted, at what level (NPC statute versus State Council regulation), and with what relation to the "interim" measures.

## 8. Current State of Knowledge
As of 2026-10-18 the instruments listed in section 5 are the operative regime among those reviewed; the list may not be exhaustive of all CAC and sectoral AI-related rules. No instrument among them was found revoked or replaced in the searches run. The Interim Measures remain titled interim. A draft comprehensive AI law has not, in sources reviewed, been published. Comparative claims about enforcement intensity are Preliminary.

## 9. Future Trajectory & Strategic Outlook
Observed: sustained use of targeted measures and amendment of existing statutes. The State Council 2026 legislative plan mentions comprehensive AI legislation but, in the reports read, gives no date for a draft or adoption. No forecast is registered. See `TIME-001` for updates.

## 10. Related Entries & Knowledge Graph
- **Foundational Concepts**:
  - [[GOV-LAW-001] Global AI Governance & Regulation](../vol08_safety_governance/GOV-LAW-001_ai_governance.md)
- **Comparative regimes**:
  - [[GOV-EU-001] The EU AI Act](../vol08_safety_governance/GOV-EU-001_the_eu_ai_act.md)
  - [[GOV-US-001] United States AI Policy](../vol08_safety_governance/GOV-US-001_united_states_ai_policy.md)
- **Related**:
  - [[GOV-INTL-001] International AI Governance and Compute Governance](../vol08_safety_governance/GOV-INTL-001_international_ai_governance_and_compute_governance.md): China's role in multilateral processes.
  - [[GOV-EXPORT-001] Export Controls and AI Geopolitics](../vol08_safety_governance/GOV-EXPORT-001_export_controls_and_ai_geopolitics.md): the external pressure on China's compute supply.

## 11. References & Foundational Literature
1. Cyberspace Administration of China et al. (2023). *生成式人工智能服务管理暂行办法 (Interim Measures for the Management of Generative AI Services)*, Order No. 15, signed 10 July 2023. https://www.cac.gov.cn/2023-07/13/c_1690898327029107.htm. [Read in Chinese: issuing bodies, effective date, Articles 2, 4, 7, 17, 20].
2. Cyberspace Administration of China, Ministry of Industry and Information Technology & Ministry of Public Security. (2022). *互联网信息服务深度合成管理规定 (Provisions on the Administration of Deep Synthesis of Internet Information Services)*, Order No. 12, dated 25 November 2022. https://www.cac.gov.cn/2022-12/11/c_1672221949354811.htm. [Read in Chinese: Order No., date, effective 10 January 2023, Article 9].
3. Cyberspace Administration of China et al. (2021). *互联网信息服务算法推荐管理规定 (Provisions on the Administration of Algorithmic Recommendations in Internet Information Services)*, Order No. 9, dated 31 December 2021. https://www.cac.gov.cn/2022-01/04/c_1642894606364259.htm. [Read in Chinese: effective 1 March 2022, filing provision].
4. Xinhua. (2025, March 14). 四部门联合发布《人工智能生成合成内容标识办法》 (source: China Cyberspace). https://www.news.cn/politics/20250314/b7a24028f2924b7681e6ed1bfbd8fade/c.html. [Official-media reproduction of a CAC statement; the CAC original of the Measures was not opened. Implicit-label contents are from the Xinhua-hosted People's Posts and Telecommunications report, https://www.news.cn/tech/20250319/a572ec4366f149c1b0b04a07ebf4931b/c.html].
5. Xinhua. (2025b, October 28). 新华鲜报｜引导AI发展！网络安全法完成修改. https://www.news.cn/politics/20251028/b3400522666a4b3f8c844b9fa6c31131/c.html. [Adoption date and 1 January 2026 effect]. Text of new Article 20 read in a reproduction citing the NPC website: https://www.secrss.com/articles/84444 (Tier 4; not the NPC original).
6. Xinhua. (2026, May 11). 国务院办公厅印发《国务院2026年度立法工作计划》. https://www.news.cn/20260511/64fb4f5178b042e6aa16bf55932f038e/c.html. [Statement on comprehensive AI legislation]. NPC Standing Committee plan wording is taken from Law Daily interview text on https://www.chinanews.com.cn/gn/2026/07-03/10651809.shtml (Tier 4), and the March 2026 work report from Xinhua, https://www.news.cn/politics/20260315/bfc2a93919af46208459d9681242158c/c.html.

## 12. Evidential Classification Matrix
| Section / Claim | Evidence Category | Supporting Source | Confidence / Status |
| :--- | :--- | :--- | :--- |
| Generative AI Interim Measures dates and articles | Established | Ref 1 (official text) | High |
| Deep synthesis provisions dates | Established | Ref 2 (official text) | High |
| Labelling measures and standard | Well Supported | Ref 4 (official-media reproduction) | Moderate-High; standard number secondary |
| Cybersecurity Law AI amendment | Well Supported | Ref 5 | Moderate-High (NPC page not opened) |
| Algorithm recommendation date | Established | Ref 3 (official text) | High |
| No comprehensive AI law enacted; planned | Well Supported (negative finding) | Ref 6 | Moderate |
| Comparative strictness | Contested | none | n/a |

## 13. Revision History
- **v0.1.0** (2026-10-15): Initial draft. Citations checked against official pages where marked; not independently fact-checked.
- **v0.2.0** (2026-10-18): Evidence audit (docs/audits/GOV-CN-001_audit.md). Algorithm provisions verified against CAC text (Order No. 9); deep synthesis and Interim Measures checked against CAC text; labelling and Cybersecurity Law amendment checked against Xinhua; 2026 legislative-plan section added; Wikipedia reference removed.