Appendixes

• Appendix A: Implementation Considerations
  • Access Controls
  • Sandboxing
  • Limiting Execution Depth and Incorporating Human Oversight
  • Requiring Strong Authentication, Signed Data Exchange, and Integrity Checks Between Model, Retriever, and Tools
  • Memory Management, Delete When Able, Encrypt or Purge Stored Prompts and Results
  • Data Provenance
  • Recording Agent and Model Activity, Tool Invocation, and Prompt Chains and Analyzing for Dual-Use Patterns
  • Supply Chain Security
  • Confidential Computing, Secure Enclaves, and Attestation
  • AI-Assisted Red-Teaming
  • Misuse-Resilience Testing
• Appendix B: Extended Security Control Tables
  • Design
  • Develop
  • Deploy
  • Operate
• Appendix C: Security Control Sources
• Appendix D: Key Sources and Recommended Reading
• Appendix E: Guide Methodology

Note: Reference links in the guide and appendixes were accurate as of the date of publication.