Review & Assurance
Review & Assurance Pack
The council's review instruments, including impact assessments, model cards, security review, red-teaming, and human oversight.
The Review & Assurance Pack provides the instruments the council uses during structured reviews. These are the artifacts that transform a committee discussion into an evidence-based decision.
In This Section
- Impact Assessments: Structured evaluation of an AI system's potential effects
- Model Cards and Datasheets: Standardized documentation for models and datasets
- Security Review: AI-specific security considerations aligned to SAIF, OWASP, and NIST
- Red-Teaming: Structured adversarial testing for AI systems
- Human Oversight: Ensuring meaningful human control proportionate to risk
Sources
This pack draws on Microsoft's AI Impact Assessment template, NIST's Generative AI Profile, Google's Secure AI Framework, OWASP's Top 10 for LLM Applications, NIST SP 800-218A, model card and datasheet standards, and AI Verify's testing framework.