NIST AI RMF¶
NIST Artificial Intelligence Risk Management Framework
Overview¶
The NIST AI Risk Management Framework (AI RMF 1.0) provides voluntary guidance for managing risks associated with AI systems throughout their lifecycle. It is organized around four core functions -- Govern, Map, Measure, and Manage -- that together form a structured approach to AI risk management.
Core Functions¶
- Govern -- establish policies, processes, and accountability structures for AI risk management
- Map -- contextualize risks by understanding the AI system's purpose, stakeholders, and operating environment
- Measure -- employ quantitative and qualitative methods to assess identified risks
- Manage -- prioritize and act on risks according to their assessed impact and likelihood
Key Topics¶
- Trustworthy AI characteristics -- validity, reliability, safety, fairness, transparency, explainability, privacy, and security
- Risk tiering -- categorization of AI systems by risk level to allocate management effort
- Stakeholder engagement -- involving affected parties in risk identification and evaluation
- Documentation and reporting -- maintaining records of risk assessments and mitigation actions
Relevance to vnvspec¶
vnvspec's requirement and evidence models map well to the Measure and Manage functions of the AI RMF:
from vnvspec import Requirement
req = Requirement(
id="REQ-FAIR-001",
statement="The model shall achieve equalized odds across demographic groups with disparity below 0.05.",
verification_method="test",
standards={"nist_ai_rmf": ["MEASURE 2.6", "MANAGE 3.1"]},
)
The coverage_report() utility helps demonstrate that all identified risks have corresponding verification evidence, supporting the Manage function's documentation requirements.
See also: vnvspec.core.trace.coverage_report for traceability coverage analysis.