Provisions on the Management of Algorithmic Recommendations

Yes. Provisions on the Management of Algorithmic Recommendations addresses Risk Assessment through 1 provision.

Does Provisions on the Management of Algorithmic Recommendations require Risk Assessment?

China • enforcing

Yes — 1 provision

This regulation imposes 4 specific requirements for Risk Assessment across 1 provision:

Periodic review — Must periodically review, evaluate, and verify algorithms, models, data, and outcomes (Art. 8)
Security monitoring — Must monitor for security risks, maintain incident response capability, and keep an illegal/undesirable-content feature database with reporting to authorities (Art. 9)
User controls — Users must be able to turn off algorithmic recommendations (Art. 17)
Security assessment — Providers with public opinion properties or social mobilization capabilities must file and undergo security assessment (Art. 24)

Obligation:

enforcing

Effective:

Mar 1, 2022

Risk tier:

all

Scope:

providers

Requirement	Details
Periodic review	Must periodically review, evaluate, and verify algorithms, models, data, and outcomes (Art. 8)
Security monitoring	Must monitor for security risks, maintain incident response capability, and keep an illegal/undesirable-content feature database with reporting to authorities (Art. 9)
User controls	Users must be able to turn off algorithmic recommendations (Art. 17)
Security assessment	Providers with public opinion properties or social mobilization capabilities must file and undergo security assessment (Art. 24)

Violation	Fine
Filing/algorithm violations	Warning, circulated criticism, suspension of updates; fine RMB 10,000-100,000 (Art. 33)