This page summarizes the Stanford DigiChina English translation of China's Internet Information Service Algorithmic Recommendation Management Provisions, a regulation governing algorithmic recommendation in internet information services. The Provisions were jointly issued on December 31, 2021 and took effect March 1, 2022. See legislation page for the structured provisions.
Provenance
- Promulgators: CAC (State Internet Information Office), MIIT, Ministry of Public Security, and SAMR, as Joint Order No. 9
- Promulgation date: 2021-12-31
- Effective date: 2022-03-01
- Translation: Rogier Creemers, Graham Webster, and Helen Toner (Stanford DigiChina), January 2022
- URL: https://digichina.stanford.edu/work/translation-internet-information-service-algorithmic-recommendation-management-provisions-effective-march-1-2022/
Summary of provisions
The Provisions are organized into five chapters.
Chapter I sets scope and assigns multi-agency coordination, with the CAC holding overall responsibility and MIIT, MPS, and SAMR handling their respective domains. It establishes a core values requirement.
Chapter II (Service Norms) requires service providers to uphold "mainstream value orientations" and "positive energy," carries a primary security responsibility, and mandates periodic review of models and data and labeling of AI-generated content. It bars addiction-inducing models, restricts user tagging (Art. 10), requires mainstream-values presentation on front pages and in hot-search rankings (Art. 11), prohibits fake news (Art. 13), prohibits rank manipulation, bot accounts, and hot-search control (Art. 14), and bars monopolistic or unfair-competition acts (Art. 15).
Chapter III (User Rights) provides for algorithm disclosure, opt-out of personalization, and tag choice and deletion. It includes minor protections against addiction-inducing pushes, elderly fraud protections, and worker protections for gig-economy dispatch algorithms (Art. 20), and it bars algorithmic price discrimination against consumers (Art. 21).
Chapter IV (Supervision) establishes a graded-categorized algorithm security management system, requires algorithm filing within 10 working days for services with public-opinion or social-mobilization capacity, and provides for security assessments (Art. 27).
Chapter V (Liability) sets fines of 10,000–100,000 yuan, along with suspension and criminal referral.
Key claims
- The Provisions have been characterized as the first comprehensive algorithm-governance law globally to include a mandatory algorithm-filing regime. (high)
- They introduce the "public opinion properties or social mobilization capacity" trigger, the filing and security-assessment gate later reused by China — Provisions on the Administration of Deep Synthesis Internet Information Services and China — Interim Measures for the Management of Generative AI Services. (high)
- They bar algorithmic price discrimination, addictive-design dark patterns, and fake-news generation (Arts. 8, 13, 21). (high)
- They oblige worker-rights protections in gig dispatch algorithms (Art. 20). (high)
The filing mechanism is the foundation of China's layered AI regulatory stack, providing the gate that later regulations such as the deep synthesis provisions and the generative-AI interim measures reuse. The consumer-protection provisions on price discrimination, algorithmic transparency, and opt-out parallel the consumer-rights framings of California CCPA Regulations (Title 11, Division 6) and Colorado AI Act (SB 24-205) and SB 25B-004 (Date Amendment). The fake-news and mainstream-values provisions have no direct Western analogue and function as content-regime instruments.
Relationships
- supports: legislation page
- depends-on: (none)
- related: China — Provisions on the Administration of Deep Synthesis Internet Information Services (reuses filing regime)
- related: China — Interim Measures for the Management of Generative AI Services (reuses filing regime)
- related: Algorithmic Pricing and Antitrust
- related: AI and Content Moderation
- related: AI and Authoritarianism
- related: AI Sovereignty