China: National Standard of the People's Republic of China: Cybersecurity Technology - Basic Safety Requirements for Generative Artificial Intelligence Services

1. Appendix A (for Reference) Main Safety Risks of Training Data and Generated
2. 1 Scope
3. Appendix A, and reference methods for safety assessment of GenAI services are provided in Appendix
4. 2 Normative Reference Documents
5. 3 Terminology and Definitions
6. 3.1 Generative Artificial Intelligence Service
7. 3.2 Service Provider
8. 3.3 Classification Model
9. 3.4 Training Data
10. 3.5 Generative Artificial Intelligence Data Annotation
11. 3.6 Functional Data Annotation
12. 3.7 Safety Data Annotation
13. 4 Training Data Safety Requirements
14. 4.1 Data Source Safety
15. 4.1.1 Data Source Selection
16. 4.1.2 Matching of Training Data from Different Sources
17. 4.1.3 Training Data Source Management and Traceability
18. 4.2 Data Content Management
19. 4.2.1 Training Data Content Filtering
20. 4.2.2 Intellectual Property Protection
21. 4.2.3 Personal Information Protection
22. 4.3 Data Annotation Safety
23. 4.3.1 Annotator Management
24. 4.3.2 Annotation Rules
25. 4.3.3 Accuracy of Annotated Content
26. 4.3.4 Isolated Storage of Annotated Data
27. 5 Model Safety Requirements
28. 5.1 Model Training Safety
29. 5.2 Model Output Safety
30. 5.3 Model Monitoring and Evaluation
31. 5.4 Model Update and Upgrade Safety
32. 5.5 Model Environment Security
33. 6 Safety Measure Requirements
34. 6.1 Applicable Service User Groups, Scenarios, and Purposes
35. 6.2 Service Transparency
36. 6.3 Collecting User-Entered Information for Use in Training
37. 6.4 Acceptance of Complaints and Reports from the Public or Users
38. 6.5 Provision of Services to Users
39. 6.6 Service Stability and Continuity
40. 6.7 On-Device Model Services
41. Appendix A
42. Appendix B: Safety Assessment Reference Methods
43. Appendix A.: B.2.1.3.2.3 Result Judgment

The outline comes from headings detected in the text, so it might miss some or include extras.