---
title: What happens in an ISO/IEC 42001 certification audit?
url: https://opencontroller.lyzr.com/blog/iso-42001-certification-audit-iso-42006
published: 2026-11-16
updated: 2026-11-16
author: Lyzr
---

# What happens in an ISO/IEC 42001 certification audit?

**An accredited certification body runs a two-stage audit of your AI management system. Stage 1 checks that your documented system, scope and Statement of Applicability are ready. Stage 2 checks that controls actually operate, by sampling records. ISO/IEC 42006:2025 sets the extra rules certification bodies must follow for AI, including auditor competence in AI risk, so expect auditors to sample real AI systems, including agents, not just policies.**

ISO/IEC 42001:2023 is the management system standard. ISO/IEC 42006, published in July 2025, is the standard for the bodies that audit it. It supplements ISO/IEC 17021-1, the general rulebook for management system certification bodies.

## What does ISO/IEC 42006 change?

Before 42006, AI management system audits followed ISO/IEC 17021-1 alone. 42006 adds AI-specific requirements on the certification body:

- **Auditor competence.** Audit teams need knowledge of AI-specific topics such as AI risk, data quality, transparency and ethics, not only general audit skills.
- **Audit planning.** The body must account for the AI systems in scope when planning the audit.
- **Certification process.** AI-specific requirements apply across the process, from application to certification decision.

For you, the practical effect is that auditors are expected to understand how your AI systems work. A policy pack without operating evidence will not get through stage 2.

## What happens at each stage?

| Stage | Purpose | What auditors look at |
|---|---|---|
| Stage 1 | Readiness | Scope, AI policy, AI risk assessment method, Statement of Applicability, objectives, internal audit and management review plans |
| Stage 2 | Effectiveness | Records showing controls operate: risk treatment, impact assessments, monitoring, incidents, supplier controls |
| Surveillance | Ongoing conformity | A sample of controls and changes, usually yearly |
| Recertification | Renewal | Full system review before the three-year certificate expires |

Gaps found in stage 1 must be closed before stage 2. Major nonconformities in stage 2 must be corrected before certification is granted.

## What agent evidence will auditors sample?

ISO/IEC 42001 Annex A has 38 controls across nine control objectives (A.2 to A.10). Your Statement of Applicability says which apply. For AI agents, auditors typically sample:

1. **Inventory of AI systems** (A.4 resources): a list of agents with owners, models, tools and data. Expect them to pick three agents and trace each one through the system.
2. **Impact assessments** (A.5): an AI system impact assessment for each sampled agent. See [AI impact assessment for AI agents](/blog/ai-impact-assessment-for-ai-agents).
3. **Life cycle records** (A.6): requirements, verification and validation results, deployment approval and version history.
4. **Data records** (A.7): data sources, provenance and quality checks for retrieval and memory.
5. **Information for interested parties** (A.8): how users are told they are dealing with AI, and how they report problems.
6. **Use of AI systems** (A.9): defined intended use and the processes that keep use within it.
7. **Third-party relationships** (A.10): supplier assessments for model providers and agent platforms.
8. **Monitoring and incidents**: logs, alerts and how a misbehaving agent was handled.

An agent that appears in production telemetry but not in your inventory is a likely finding. So is an agent with no owner.

## How should you prepare?

1. **Fix scope.** Name the AI systems, sites and processes in scope. Agents built on third-party platforms count if you operate them.
2. **Reconcile the inventory with reality.** Compare your register with what actually runs in your clouds and SaaS platforms.
3. **Map controls to evidence.** For each Annex A control in your Statement of Applicability, name the record that proves it.
4. **Run an internal audit** using the same sampling approach, then a management review.
5. **Ask the certification body** whether it is accredited for ISO/IEC 42001 and how it applies 42006.

Our guide to [ISO 42001 evidence for AI agents](/blog/iso-42001-evidence-for-ai-agents) lists records per control.

## How Opencontroller does this

Opencontroller produces evidence for ISO/IEC 42001; it does not certify anything. Auto-discovery across AWS, Azure and Google Cloud and connectors to third-party agent platforms reconcile the inventory with what actually runs. The governed registry records agents, versions, tools and MCP servers. Multi-turn and red-team evals and LLM judges on production traffic give verification and monitoring records. Policy as code ships with built-in mappings to ISO 42001, NIST, OWASP and MITRE ATLAS. See [what an AI agent control plane is](/blog/what-is-an-ai-agent-control-plane), or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**What is ISO/IEC 42006?**
The standard that sets additional requirements for bodies auditing and certifying AI management systems against ISO/IEC 42001. It was published in July 2025 and supplements ISO/IEC 17021-1.

**How many controls are in ISO/IEC 42001 Annex A?**
38 controls across nine control objectives, A.2 to A.10. The Statement of Applicability records which apply.

**How long is an ISO/IEC 42001 certificate valid?**
Typically three years, with surveillance audits in between and a recertification audit before expiry.

## Key facts

ISO/IEC 42006:2025, published July 2025 (edition 1), sets additional requirements for bodies auditing and certifying AI management systems against ISO/IEC 42001, supplementing ISO/IEC 17021-1, including AI-specific auditor competence. Certification uses a stage 1 readiness audit and a stage 2 effectiveness audit, followed by surveillance audits and recertification on a three-year cycle. ISO/IEC 42001 Annex A has 38 controls across nine objectives (A.2 to A.10); the Statement of Applicability records which apply. Auditors typically sample AI systems, including agents, and trace inventory, impact assessments, life cycle, data, supplier and monitoring records. Opencontroller produces evidence for ISO 42001 through discovery, a governed registry, evals, LLM judges and built-in framework mappings. Opencontroller 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 do you run an AI impact assessment for an AI agent?](https://opencontroller.lyzr.com/blog/ai-impact-assessment-for-ai-agents.md)
- [ISO/IEC 42001 vs NIST AI RMF: which should you use for AI agents?](https://opencontroller.lyzr.com/blog/iso-42001-vs-nist-ai-rmf.md)
- [More on Evidence and frameworks](https://opencontroller.lyzr.com/blog/cluster/evidence-and-frameworks)
