AI Governance With the NIST AI RMF and ISO/IEC 42001: A Practical Starting Point

Most organizations can start AI governance with five steps — inventory AI use, set an acceptable-use policy, assess risk by use case, manage third-party AI risk and assign oversight. The NIST AI Risk Management Framework (Govern, Map, Measure, Manage) provides a flexible structure, while ISO/IEC 42001 defines a certifiable AI management system.

AI adoption has outpaced AI governance in most organizations. Employees use generative AI tools daily, vendors are adding AI features to software you already license, and customers are starting to ask how you manage AI risk. The good news: you don't need a massive program to get started. You need a clear structure and a few practical controls.

Start with the risk you actually have

For most organizations, AI risk today comes from using AI rather than building it:

  • Employees pasting confidential data into public generative AI tools
  • AI features enabled by default in productivity, CRM, HR or finance software
  • AI-generated content or code used without review
  • Vendors using your data to train their models
  • Decisions influenced by AI outputs that no one validated

A governance program should address these first.

Five practical first steps

1. Inventory AI use. Survey business units, review software contracts and check which applications have AI features enabled. You can't govern what you can't see.

2. Publish an acceptable-use policy. Define approved tools, what data can and can't be used with them, review expectations for AI outputs and how to request new tools. A clear policy enables adoption as much as it restricts it.

3. Assess risk by use case. Not all AI use carries the same risk. Drafting marketing copy is different from screening job applicants or summarizing patient records. Tier use cases by data sensitivity, impact on individuals and regulatory exposure.

4. Manage third-party AI risk. Add AI questions to vendor due diligence — how is our data used, is it used for training, where is it processed, how are outputs validated — and address them in contracts.

5. Assign oversight. Establish an executive sponsor and a cross-functional group to approve higher-risk use cases, monitor the program and report to leadership.

The NIST AI Risk Management Framework

The NIST AI RMF 1.0, released in January 2023, is a voluntary framework for managing risks to individuals, organizations and society from AI systems. It describes characteristics of trustworthy AI — such as being valid and reliable, safe, secure and resilient, accountable and transparent, explainable, privacy-enhanced and fair — and organizes risk management into four functions:

  • Govern — cultivate a culture of risk management with policies, accountability and oversight
  • Map — understand context and identify risks for each AI system
  • Measure — analyze, assess and track identified risks
  • Manage — prioritize and act on risks, and monitor over time

NIST has also published a Generative AI Profile (NIST AI 600-1) that describes risks specific to generative AI and suggested actions.

ISO/IEC 42001

ISO/IEC 42001:2023 specifies requirements for an AI management system (AIMS) — the policies, processes and controls an organization uses to develop, provide or use AI responsibly. Like ISO 27001, it follows the familiar management system structure (context, leadership, planning, support, operation, performance evaluation and improvement) and organizations can be certified by an accredited certification body.

NIST AI RMF or ISO/IEC 42001?

NIST AI RMF ISO/IEC 42001
Type Voluntary framework Management system standard
Certification No Yes, by accredited bodies
Strength Flexible, detailed risk guidance Formal, auditable structure
Best for Building and maturing a program Demonstrating assurance to customers

The two work well together: many organizations use the AI RMF's guidance to design their approach and ISO/IEC 42001's structure to formalize and certify it.

Connect AI to the programs you already have

AI governance shouldn't be a silo. Fold AI risks into your enterprise risk assessment, add AI considerations to security and privacy reviews, extend vendor risk management to AI providers and include AI controls in your internal audit plan. That integration is what turns a policy into a program.

The bottom line

Start small and practical: know where AI is used, set clear rules, focus scrutiny on high-risk use cases and manage your vendors. Then use the NIST AI RMF and ISO/IEC 42001 to mature the program into something you can measure, audit and — if customers require it — certify.

Frequently asked questions

Is the NIST AI RMF mandatory?

No. The AI RMF is voluntary guidance. It is widely referenced, however, and aligning with it helps demonstrate responsible practices to customers, regulators and auditors.

Should we pursue ISO/IEC 42001 certification?

Consider it if customers ask for independent assurance over your AI practices — especially if you provide AI-based products — or if you already maintain ISO 27001 and want an integrated approach.

Who should own AI governance?

Ownership varies, but effective programs typically have an executive sponsor and a cross-functional committee including legal, privacy, security, IT, risk and business leaders.

This article is for general information only and does not constitute legal, accounting or other professional advice. Requirements change; consult the cited sources and a qualified advisor about your specific situation.

Formos Consulting Risk, compliance, cybersecurity and transformation advisors

Keep reading

Related insights

All insights

Let’s talk about what you need.

Tell us where you are and where you need to be. A senior member of our team will follow up to talk through options.