AI Regulation ReferenceDoes Interim Measures for Generative AI Services require Risk Assessment?
China • enforcing
Yes — 1 provision
Requirements at a glance
This regulation imposes 5 specific requirements for Risk Assessment across 1 provision:
- Security assessment — Services with public opinion properties require security assessment
- Algorithm filing — Must file algorithms with regulators per Algorithm Recommendation Regulations
- Training data compliance — Training data must be lawfully obtained and not infringe IP or personal data
- Model monitoring — Must monitor outputs and stop illegal content generation, delete, report
- Cooperation with inspections — Must cooperate with regulatory inspections
Generative AI Risk Assessment #
Requirements
| Requirement | Details |
|---|---|
| Security assessment | Services with public opinion properties require security assessment |
| Algorithm filing | Must file algorithms with regulators per Algorithm Recommendation Regulations |
| Training data compliance | Training data must be lawfully obtained and not infringe IP or personal data |
| Model monitoring | Must monitor outputs and stop illegal content generation, delete, report |
| Cooperation with inspections | Must cooperate with regulatory inspections |
Penalties
| Violation | Fine |
|---|---|
| Non-compliance | Warnings, corrections, service suspension; criminal liability for severe violations |