NIST AI Risk Management Framework

The NIST AI Risk Management Framework (AI RMF 1.0), released on 26 January 2023, is a voluntary guidance document developed by the National Institute of Standards and Technology to help organisations design, develop, deploy, and use AI systems in a trustworthy and responsible manner. While not legally binding, the AI RMF has become a de facto standard for AI governance in the United States and is increasingly referenced in regulatory guidance worldwide.

The framework is structured around four core functions: Govern, Map, Measure, and Manage. These functions provide a flexible, structured approach to AI risk management that can be adapted to any organisation's size, sector, and risk tolerance.

The GOVERN function establishes and maintains the organisational structures, policies, and processes for AI risk management. It emphasises leadership accountability, stakeholder engagement, and the integration of AI risk management into broader enterprise risk management. This function recognises that effective AI governance requires cultural change, not just technical controls.

The MAP function is about understanding the context in which AI systems operate. It involves identifying and categorising AI systems, understanding their intended purposes and potential impacts, and recognising the broader societal context of AI deployment. Mapping also includes identifying relevant stakeholders and understanding the legal and regulatory landscape.

The MEASURE function focuses on employing quantitative and qualitative methods to analyse, assess, benchmark, and monitor AI risks and their related impacts. This includes developing metrics for trustworthiness characteristics such as accuracy, fairness, privacy, security, resilience, transparency, explainability, and accountability.

The MANAGE function involves allocating resources and implementing plans to respond to, recover from, and communicate about AI risks. It includes prioritising risks, implementing mitigation strategies, and establishing processes for ongoing monitoring and adjustment.

The companion document, the NIST AI RMF Playbook, provides suggested actions and references for each subcategory, making the framework highly practical for implementation. The Playbook is a living document that NIST updates as practices evolve.

The AI RMF was developed through an extensive multi-stakeholder process involving hundreds of organisations from industry, academia, civil society, and government. This collaborative development process has given the framework broad legitimacy and acceptance across sectors.

In the context of the US regulatory landscape, the AI RMF serves as a foundational reference. The 2023 Executive Order on AI Safety explicitly references the NIST AI RMF, and several US federal agencies have incorporated it into their AI governance guidance. State-level AI legislation, such as the Colorado AI Act, also references NIST standards.