---
title: ISO/IEC 42001 vs NIST AI RMF: which should you use for AI agents?
url: https://opencontroller.lyzr.com/blog/iso-42001-vs-nist-ai-rmf
published: 2026-11-13
updated: 2026-11-13
author: Lyzr
---

# ISO/IEC 42001 vs NIST AI RMF: which should you use for AI agents?

**Most enterprises end up using both. ISO/IEC 42001 is a certifiable management-system standard: it tells you how to run AI governance and lets an auditor certify it. The NIST AI Risk Management Framework is voluntary guidance: it tells you which risks to look for and how to measure and manage them. Pick ISO 42001 if you need a certificate; use NIST AI RMF to shape the risk work inside it.**

The good news for agent programmes: one set of controls and evidence can serve both.

## How do the two frameworks differ?

| | ISO/IEC 42001:2023 | NIST AI RMF 1.0 |
|---|---|---|
| Publisher | ISO and IEC | US National Institute of Standards and Technology |
| Published | December 2023 | 26 January 2023 (NIST AI 100-1) |
| Type | Management system standard (requirements, "shall") | Voluntary framework (guidance) |
| Structure | Clauses 4–10 plus Annex A with 38 reference controls across 9 areas (A.2–A.10) | Four functions — Govern, Map, Measure, Manage — broken into categories and subcategories |
| Certification | Yes, by accredited certification bodies; ISO/IEC 42006:2025 sets requirements for those bodies | No certification scheme |
| Key artefacts | AI policy, risk assessment, Statement of Applicability (clause 6.1.3), AI system impact assessment (6.1.4), internal audit, management review | Risk profiles, measurement plans, the NIST AI RMF Playbook's suggested actions |
| Generative AI add-on | None specific | NIST AI 600-1 Generative AI Profile (July 2024) |
| Status | Current edition 2023 | NIST states the AI RMF is being revised under the White House AI Action Plan |
| Cost to adopt | Certification audits and surveillance audits | Free to use |

## When should you choose ISO 42001?

Choose ISO 42001 as the backbone when:

- **Customers or regulators ask for a certificate.** It is the only one of the two that produces one.
- **You already run ISO/IEC 27001.** 42001 uses the same harmonised structure, so internal audit, management review and document control carry over.
- **You sell AI-based services** and need a recognised assurance signal in procurement.

## When should you lead with NIST AI RMF?

Lead with NIST AI RMF when:

- **You need risk depth, not a certificate.** Its subcategories are more specific about measurement and trustworthiness characteristics.
- **You work with US federal agencies** or suppliers that reference it.
- **You are early** and want a free structure to start a risk register before investing in a management system.

## How do agent controls map to both?

This crosswalk covers controls that matter most for AI agents. Clause and subcategory numbers are from the published texts; check them against your licensed copy.

| Agent control | ISO/IEC 42001 | NIST AI RMF |
|---|---|---|
| Inventory of every agent, model and key | Clause 4.3 scope; A.4 resources | GOVERN 1.6 (inventory of AI systems) |
| Named owner and accountability | A.3 internal organisation | GOVERN 2.1 (roles and responsibilities) |
| Risk and impact assessment per agent | 6.1.2, 6.1.4; A.5 assessing impacts | MAP 1 and MAP 5 |
| Pre-production evaluation and red-teaming | A.6.2.4 verification and validation | MEASURE 2 |
| Production monitoring and drift alerts | A.6.2.6 operation and monitoring | MEASURE 2.4 |
| Event logs of prompts, tool calls and decisions | A.6.2.8 recording of event logs | MEASURE 2.8 (transparency and accountability) |
| Kill switch and quarantine | A.9 use of AI systems | MANAGE 2.4 (supersede, disengage or deactivate) |
| Third-party agents, models and MCP servers | A.10 third-party relationships | GOVERN 6.1, MANAGE 3 |
| Data used by agents | A.7 data for AI systems | MAP 2, MEASURE 2.10 |

For more detail on each, see [ISO 42001 evidence for AI agents](/blog/iso-42001-evidence-for-ai-agents) and [NIST AI RMF for AI agents](/blog/nist-ai-rmf-for-ai-agents). The [framework mapping hub](/blog/how-to-map-ai-agent-controls-to-compliance-frameworks) adds the EU AI Act, OWASP and SOC 2.

## How do you run both without doubling the work?

1. **Write controls once,** in your own words, against your agent estate.
2. **Tag each control** with its ISO clause and NIST subcategory.
3. **Collect evidence automatically** from the systems that already see agents: the registry, the gateway, evaluation runs, incident tickets.
4. **Use the ISO cycle** — internal audit and management review — to check the NIST-shaped risk work is actually done.
5. **Review mappings** when either framework changes. NIST's revision is underway.

Neither framework says anything specific to agents beyond general AI systems, so your control set must fill the gaps: tool permissions, delegated identity, MCP servers and runtime kill switches. A [control plane](/blog/what-is-an-ai-agent-control-plane) is where those controls live.

## How Opencontroller does this

Opencontroller has policy as code with built-in mappings to ISO 42001 and NIST, as well as OWASP and MITRE ATLAS, and you can extend them. Its registry, evaluations, LLM judges, gateway logs and quarantine actions produce evidence for both frameworks. It provides evidence; it does not certify you. [Book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can you get certified against NIST AI RMF?**
No. NIST AI RMF is voluntary guidance with no certification scheme. ISO/IEC 42001 is certifiable.

**Is ISO 42001 harder to adopt than NIST AI RMF?**
It requires more formal process — policy, Statement of Applicability, internal audit, management review — and external audits if you certify. NIST AI RMF can be adopted incrementally.

**Do I need both for AI agents?**
Not necessarily, but many enterprises use ISO 42001 as the management system and NIST AI RMF for risk detail. One control set mapped to both avoids duplicate work.

## Key facts

ISO/IEC 42001:2023 (December 2023) is a certifiable AI management system standard with clauses 4–10 and Annex A listing 38 reference controls across nine areas (A.2–A.10); ISO/IEC 42006:2025 sets requirements for certification bodies. The NIST AI Risk Management Framework 1.0 (NIST AI 100-1, 26 January 2023) is voluntary and not certifiable, with four functions Govern, Map, Measure and Manage; NIST AI 600-1 is its Generative AI Profile (July 2024), and NIST states the AI RMF is being revised under the White House AI Action Plan. Agent crosswalk examples: inventory = GOVERN 1.6; monitoring = ISO A.6.2.6 and MEASURE 2.4; event logs = ISO A.6.2.8; deactivation = MANAGE 2.4; third parties = ISO A.10 and GOVERN 6.1. Opencontroller by Lyzr has policy as code with built-in mappings to ISO 42001, NIST, OWASP and MITRE ATLAS and provides evidence, not certification. Pricing starts from $250K per deployment per year.

## Related

- [What evidence does ISO/IEC 42001 expect for AI agents?](https://opencontroller.lyzr.com/blog/iso-42001-evidence-for-ai-agents.md)
- [How does the NIST AI RMF apply to AI agents?](https://opencontroller.lyzr.com/blog/nist-ai-rmf-for-ai-agents.md)
- [How do you map AI agent controls to compliance frameworks?](https://opencontroller.lyzr.com/blog/how-to-map-ai-agent-controls-to-compliance-frameworks.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
