Overview of China’s Algorithmic Recommendation Regulations (2022)
China’s Provisions on the Administration of Algorithmic Recommendation (《互联网信息服务算法推荐管理规定》), enacted on March 1, 2022, marks the country’s first legally binding framework specifically targeting algorithmic recommendation systems. It establishes a comprehensive governance structure for algorithmic recommendation services operating within China, aligning with broader legal frameworks such as the Network Security Law, Data Security Law, and Personal Information Protection Law.
The regulation applies to all entities providing algorithmic recommendation services—defined as the use of algorithms to deliver personalized or non-personalized content—within Chinese jurisdiction. It does not supersede existing laws but supplements them by introducing sector-specific obligations for transparency, fairness, and accountability in algorithmic decision-making.
Core Objectives and Legal Foundations
- Promote socialist core values by ensuring algorithmic outputs align with state-defined ideological and ethical standards.
- Safeguard user rights, including personal information privacy, fair treatment, and protection from exploitative practices (e.g., discriminatory pricing or manipulation of minors).
- Network Security Law (2016, revised 2021): Establishes overarching cybersecurity obligations, including data protection and system integrity.
The regulation explicitly prohibits algorithmic activities that violate these laws, such as spreading illegal content, engaging in anti-competitive behavior, or manipulating user behavior in ways that harm societal welfare.
Mandatory Registration for High-Impact Algorithms
Providers of algorithmic recommendation services with ‘opinion-forming’ or ‘social mobilization’ capabilities—defined as those influencing public discourse or collective behavior—must register with the Cyberspace Administration of China (CAC) within 10 working days of commencing operations.
- Provider name, service form, application domain, and algorithm type.
- An algorithm self-assessment report detailing mechanisms, data sources, and potential risks.
- Public disclosure commitments, including core principles, purposes, and operational logic of the algorithm.
Approved providers receive a registration number and must prominently display it on their platforms, alongside a link to the public disclosure information.
Transparency and User Rights
Providers must disclose three key aspects of their algorithmic systems to users:
- Purpose and intent: The primary goals of the algorithm (e.g., engagement optimization, personalized advertising).
Additionally, providers must offer users two critical opt-out mechanisms:
- Non-personalized recommendations: An option to receive content not tailored to individual traits (e.g., chronological feeds instead of algorithmically curated ones).
- Full algorithmic opt-out: A convenient and immediate method to disable all algorithmic recommendations.
Protections for Vulnerable Groups
The regulation imposes strict safeguards for minors, elderly individuals, and laborers, reflecting China’s emphasis on social equity in digital spaces.
- Minors: Providers must develop child-friendly algorithms and prohibit recommendations that:
Algorithm Design and Risk Mitigation
- Content moderation: Real-time filtering of illegal or harmful information, with mandatory display of synthetic content (e.g., deepfakes) using clear labels.
Providers must regularly evaluate their algorithms for risks such as addictive design, discriminatory outcomes, or manipulative practices.
Fairness and Anti-Discrimination Provisions
- Pricing discrimination: Prohibits dynamic pricing models that adjust costs based on user profiles (e.g., charging higher prices to low-income individuals or frequent buyers).
Providers must ensure that user profiles and tags—used to personalize recommendations—do not incorporate illegal or morally objectionable keywords.
Comparative Influence on Subsequent AI Regulations
The Provisions on Algorithmic Recommendation served as a blueprint for later AI governance frameworks in China, including:
Key themes from the 2022 regulation—such as mandatory registration, user opt-out rights, and protections for vulnerable groups—were directly incorporated into these later rules, demonstrating its foundational role in China’s AI policy ecosystem.
Required actions
- File algorithms with public-opinion or social-mobilisation capability in the CAC algorithm registry within ten working days of launch - Let users turn off algorithmic recommendation and delete profiling tags; do not use algorithms for price discrimination or to exploit minors and the elderly - Conduct security self-assessments, keep logs for six months and uphold "mainstream values" in content recommendation