China Algorithm Recommendation Regulations

Applies To

Overview

The Provisions on the Management of Algorithmic Recommendations (互联网信息服务算法推荐管理规定) were issued as Order No. 9 on 31 December 2021 by the Cyberspace Administration of China (CAC) and three co-regulators. They were published on 4 January 2022 and entered into force on 1 March 2022. The Provisions apply to any provider of 'algorithmic recommendation services', defined broadly to encompass content recommendation, search ranking, advertising targeting, push notifications, and similar automated decision mechanisms delivered via internet information services to users located in China. Regulators manage providers by category and grade, weighing factors such as user scale. Any provider with public opinion attributes or social mobilization capability must file its algorithm with the CAC within 10 working days of launch. All in-scope providers are obligated to maintain written algorithm policy documentation, conduct regular self-assessments, and submit to government audits and security reviews. The Provisions embed user-rights obligations requiring that individuals be informed when they are subject to algorithmic recommendations, offered the ability to opt out of profiling based on personal characteristics, and given access to mechanisms to correct or delete tags (labels the system assigns to describe them) attached to their behavioral profiles. Providers are prohibited from deploying recommendation algorithms to disseminate unlawful content, to engage in price discrimination against users based on inferred characteristics, to exploit minor users, or to foster addictive usage patterns. Employers using algorithmic management tools for workers in gig-economy or platform-labor contexts must ensure algorithmic fairness and cannot use such systems to circumvent labor protections. The Provisions interact with China's Personal Information Protection Law (PIPL), Data Security Law (DSL), and the subsequently enacted Deep Synthesis Regulations (covering deepfakes and other AI-generated media) and Generative AI Measures, forming a layered domestic AI governance architecture.

Key Requirements

• •Providers must tell users conspicuously that they offer algorithmic recommendation, and publish its basic principles, purpose, and main mechanisms (Article 16).
• •Users must be given an accessible mechanism to opt out of personalized recommendation and to request non-personalized alternatives.
• •Providers must maintain transparent algorithmic policy documentation and make summaries available to users.
• •Periodic self-assessments and compliance audits of recommendation algorithms are mandatory.
• •Providers with public opinion attributes or social mobilization capability must file their algorithms with the CAC within 10 working days of launch (Article 24).
• •Algorithmic systems must not be used to create or spread disinformation, illegal content, or content that undermines social order.
• •Price discrimination based on user profiling is prohibited.
• •Algorithmic systems targeting minors must comply with enhanced protections and must not encourage addictive usage.
• •Platforms using algorithmic tools to manage gig or platform workers must disclose algorithmic rules to workers and protect lawful labor rights.
• •Providers must maintain logs and technical records sufficient to support regulatory inspection and incident investigation.

What Your Organization Must Do

• →Map every China-facing service feature that ranks, recommends, targets ads or sends push notifications.
• →Publish plain notice of algorithmic recommendation, its basic principles, purpose and main mechanisms, per Article 16.
• →Build an accessible user option to switch off personalized recommendation and correct or delete profile tags.
• →File algorithms with the Cyberspace Administration of China within 10 working days where public opinion or social mobilization features exist.
• →Schedule periodic self-assessments and keep algorithm policy documents, logs and records ready for regulatory inspection.
• →Audit pricing, minor-user and gig-worker dispatch tools for profiling-based pricing, addictive design and labor rights exposure.

Playbook Guidance

Step-by-step implementation guidance for compliance teams.

Frequently Asked Questions

Do China's Algorithm Recommendation Regulations apply to foreign companies serving users in China?

The Provisions cover algorithmic recommendation services provided within China. Foreign companies running China-facing services with recommendation features should assume they are in scope, including the notice, opt-out, and filing duties.

Which companies must register their algorithmic recommendation services with the CAC?

Any provider whose service has public opinion attributes or social mobilization capability must file within 10 working days of launch. There is no user-scale threshold; user scale only feeds regulators' graded management.

What does the price discrimination prohibition actually prohibit under these Provisions?

Providers are prohibited from using algorithmic recommendation systems to charge different prices to users based on inferred personal characteristics or behavioral profiles. This directly targets dynamic pricing practices that offer worse terms to users identified as less price-sensitive, a common concern on e-commerce and ride-hailing platforms.

What specific protections apply to minor users under the China Algorithm Recommendation Regulations?

Algorithmic systems targeting minors must comply with enhanced protections and are expressly prohibited from encouraging addictive usage patterns. Platforms must ensure recommendation logic applied to minors does not exploit behavioral vulnerabilities, and these obligations interact with China's broader minor-protection rules for online services.

What are the obligations for gig-economy platforms using algorithmic dispatch or performance management tools?

Platforms using algorithms to manage gig or platform workers must disclose the relevant algorithmic rules to workers and must not use such systems to circumvent applicable labor protections under Chinese law. Compliance counsel should audit dispatch and performance-management tools specifically for algorithmic-fairness and labor-rights exposure.

How do the Algorithm Recommendation Regulations relate to China's PIPL and Data Security Law?

The Provisions operate alongside the Personal Information Protection Law and the Data Security Law, forming a layered domestic AI governance framework. Behavioral profile tags used in recommendation systems must be correctable and deletable per PIPL user-rights requirements, so compliance programs should address all three instruments together rather than in isolation.