AI Governance

The AI Risk Management Framework Explained

By Samaksh·2026-10-05·12 min read

Short answer: The AI Risk Management Framework (AI RMF 1.0), published by NIST in January 2023, gives organizations a structured, voluntary way to identify, assess, and manage AI risks across the full system lifecycle through four core functions: Govern, Map, Measure, and Manage.

TL;DR: NIST's AI RMF is the most widely adopted AI risk framework in use today. It organizes AI risk management around four functions and a set of measurable outcomes, and it is increasingly referenced in federal procurement, sector regulations, and enterprise vendor requirements. Only 33% of organizations have formally adopted it — which means early movers have a compliance advantage.

Key Takeaways - The NIST AI RMF has four functions: Govern, Map, Measure, and Manage — each with subcategories and informative references. - Govern is the foundation: it establishes the policies, roles, and culture that make the other three functions work. - The framework is voluntary but is being incorporated into federal procurement and sector-specific regulations. - 33% of organizations report using the NIST AI RMF; 36% use ISO/IEC 42001 — adoption is growing but still far below AI deployment rates. - Applying the framework starts with cataloguing your AI systems, not rewriting your entire risk program.

What the NIST AI RMF is and where it comes from

The NIST AI Risk Management Framework (AI RMF 1.0) was published by the U.S. National Institute of Standards and Technology in January 2023. It was developed through an open, multi-stakeholder process that involved hundreds of organizations across industry, government, and academia.

The framework is explicitly voluntary and designed to be sector-agnostic. It does not prescribe specific technologies or point solutions. Instead, it defines a vocabulary, a set of outcomes, and a structured process that organizations can adapt to their own context, risk appetite, and existing governance structures.

The AI RMF is distinct from, but complementary to, the NIST Cybersecurity Framework. Many organizations use both: the CSF for cybersecurity risk and the AI RMF for AI-specific risk. The two share a similar functional structure, which makes integration relatively straightforward.

Who should use it

The AI RMF is relevant to any organization that: - Develops AI systems for internal use or external deployment - Purchases or integrates third-party AI components - Operates AI systems that affect individuals or critical processes - Faces regulatory or contractual requirements to demonstrate responsible AI practices

For compliance and risk professionals, the framework provides a structured basis for AI risk assessment, audit readiness, and reporting. For security teams, it maps AI-specific risks to existing controls frameworks.

The four core functions

The framework organizes AI risk management into four functions. They are not strictly sequential — Govern runs continuously across all of them — but they build on each other logically.

Govern

Govern is the foundation. It establishes the organizational context: the policies, processes, roles, and culture that enable effective AI risk management. Without Govern, the other three functions have no authority, resources, or accountability structure to operate within.

Key outcomes in Govern include: - Documented AI risk policies and acceptable use standards - Defined roles and responsibilities for AI risk ownership - Processes for escalating and resolving AI risk issues - A culture that treats AI risk as a shared organizational concern, not just a technical one

Govern also requires alignment with broader organizational risk management — connecting AI risk to enterprise risk, legal and compliance, and strategic planning.

Map

Map is the discovery and context-setting function. Before you can assess or treat AI risks, you need to know what AI systems you have, what they do, and what harms they could cause.

Mapping activities include: - Cataloguing AI systems across the organization - Defining the intended purpose and context of use for each system - Identifying affected stakeholders — including those who may be harmed - Characterizing the AI system's inputs, outputs, and decision points - Assessing the potential for harm based on the system's context and impact scope

Map connects to third-party AI risk: if you use vendor-supplied AI components, they need to be mapped along with internally developed systems.

Measure

Measure turns the context from Map into quantified or qualified risk assessments. It asks: given what we know about this system and its context, how significant are the risks, and how well are current controls working?

Measurement activities include: - Testing AI systems for performance, bias, robustness, and security vulnerabilities - Evaluating explainability and transparency against stakeholder requirements - Assessing data quality across the AI lifecycle - Benchmarking against relevant standards and best practices - Documenting measurement results in a way that supports decision-making

This is where technical controls intersect with governance. Security testing of AI systems — adversarial robustness, prompt injection testing for generative models, model inversion checks — falls within the Measure function.

Manage

Manage is the treatment and response function. It takes the outputs of Map and Measure and applies them: prioritizing risks, implementing controls, responding to incidents, and continuously improving the program.

Key Manage activities include: - Prioritizing AI risks based on likelihood and impact - Implementing and documenting risk treatments (accept, mitigate, transfer, avoid) - Maintaining a response and recovery capability for AI incidents - Monitoring AI systems in production for ongoing risk - Feeding lessons learned back into Govern, Map, and Measure

Manage is where the risk register lives — and for organizations building a formal AI risk register, the Manage function provides the operational context.

How the four functions fit together

Figure 1. The four NIST AI RMF functions. Govern is the continuous foundation; Map, Measure, and Manage form the operational cycle.

Function	Primary question	Key outputs
Govern	Do we have the policies, roles, and culture to manage AI risk?	AI risk policy, ownership matrix, escalation process
Map	What AI systems do we have, and what could go wrong?	AI system inventory, stakeholder harm analysis, context profiles
Measure	How significant are the risks, and how well are controls working?	Risk assessments, bias test results, explainability ratings, audit evidence
Manage	What do we do about the risks we have identified?	Risk register, treatment plans, incident response records, monitoring results

The AI RMF and other major frameworks

The AI RMF does not exist in isolation. It is designed to work alongside existing standards and regulations.

NIST AI RMF and the EU AI Act

The EU AI Act (Regulation (EU) 2024/1689) is a binding regulation with specific requirements for high-risk AI systems. The AI RMF is a voluntary framework. They are complementary: organizations using the AI RMF build the governance infrastructure that the EU AI Act's conformity assessment and technical documentation requirements demand. Using the AI RMF makes EU AI Act compliance significantly more tractable.

NIST AI RMF and ISO/IEC 42001

ISO/IEC 42001 is a certifiable AI management system standard published in 2023. It aligns closely with the AI RMF in structure and intent. For organizations that need a certifiable credential — particularly in B2B contexts where customers require it — ISO/IEC 42001 provides that. The AI RMF provides deeper prescriptive guidance on risk activities.

NIST AI RMF and existing GRC programs

If your organization already has a GRC program or an enterprise cyber risk management framework, the AI RMF extends rather than replaces it. AI risk categories feed into the enterprise risk register; AI control testing feeds into the audit program. The frameworks share a common vocabulary of risk identification, assessment, and treatment.

Adoption rates and what they tell you

AI governance survey data (2026) indicates that 33% of organizations report using the NIST AI RMF and 36% use ISO/IEC 42001. Both figures sit well below the roughly 88% of organizations actively deploying AI, confirming that formal governance continues to lag adoption. The broader adoption trend is consistent with the IBM Global AI Adoption Index and with ENISA's analysis of AI uptake outpacing security maturity in its Threat Landscape reporting.

Adoption is also highly uneven across sectors, so in most industries there is still a first-mover advantage for organizations that formalize AI risk management early.

How to start applying the AI RMF

The most common mistake organizations make is trying to implement all four functions simultaneously before they understand what AI they actually have. A more practical sequence:

Start with Govern — establish a minimal AI risk policy and assign ownership. Even a one-page policy is better than none.
Run a Map sprint — catalogue every AI system in use, including vendor-supplied tools. Use the AI RMF's AI system card template as a starting point.
Prioritize for Measure — not every system needs the same depth of assessment. Focus Measure efforts on systems with the highest potential for harm.
Build Manage incrementally — start a risk register with the top five to ten AI risks. Build response procedures for the highest-priority items first.

For compliance professionals, the AI governance training for compliance professionals post covers the skills and credentials that support framework implementation.

Frequently asked questions

Is the NIST AI RMF mandatory?

Currently it is voluntary for most organizations. However, it is being referenced in U.S. federal procurement requirements, and several sector regulators are adopting its structure as a baseline. Organizations that operate in regulated sectors or serve government customers should treat it as de facto required.

How does the AI RMF differ from the NIST Cybersecurity Framework?

The NIST CSF addresses cybersecurity risk broadly. The AI RMF addresses risks specific to AI systems — bias, opacity, data quality, emergent behavior, and the ethical dimensions of automated decision-making. Both use a functional structure and both can be operated together. The AI RMF's Govern function has no direct equivalent in the CSF.

How long does it take to implement the AI RMF?

Implementation depth varies significantly by organization size and AI maturity. In our experience training compliance teams, a minimal implementation — basic Govern policies, a system inventory, and a risk register — can typically be built in roughly 60 to 90 days, and a comprehensive, auditable program covering all four functions typically takes about 6 to 12 months plus ongoing maintenance. NIST frames the AI RMF as a continuous, iterative process rather than a fixed-duration project, so treat these as practitioner estimates rather than guarantees.

Does using the AI RMF satisfy EU AI Act requirements?

The AI RMF and EU AI Act are complementary but not identical. Organizations deploying high-risk AI systems under the EU AI Act need to meet specific conformity assessment, technical documentation, and human oversight requirements that go beyond what the AI RMF specifies. Using the AI RMF builds much of the governance infrastructure the EU AI Act requires, but a gap analysis against specific EU AI Act obligations is still necessary.

ThreatBlock's AI governance course trains compliance and risk professionals on AI governance frameworks and implementation.

About the author: Samaksh is Lead Instructor at ThreatBlock and holds the OSCP, CRTP, CRTO, and CEH certifications, with expertise in cybersecurity risk management and AI governance. Read more on the author page · Editorial policy.