What ISO/IEC 42001 actually requires
Published in December 2023, ISO/IEC 42001 defines an AI Management System (AIMS). It uses the same management-system structure as ISO/IEC 27001, so if you have run an ISMS the shape is familiar: a Plan-Do-Check-Act cycle across these clauses.
| Clause | What it asks for |
|---|---|
| 4. Context | Understand where and how AI is used, by whom, and the scope of the AIMS. |
| 5. Leadership | An AI policy, roles, and accountability set by leadership. |
| 6. Planning | AI risk assessment and AI system impact assessment, with objectives to address them. |
| 7. Support | Resources, competence, awareness, and documented information. |
| 8. Operation | Run the controls: manage the AI lifecycle, data, and third-party AI in practice. |
| 9. Performance | Monitor, measure, audit internally, and review with management. |
| 10. Improvement | Correct nonconformities and improve the system over time. |
Annex A lists the AI-specific controls (control objectives across areas such as AI policy, internal organization, resources for AI systems, the AI system lifecycle, data for AI, information for interested parties, use of AI systems, and third-party relationships). Clauses 9 and 10 are where most programs struggle, because they need continuous, current evidence rather than a one-time document.
How Senserva helps today: auditing Microsoft AI and agents
An AI policy on paper is not evidence. ISO/IEC 42001 wants proof that the controls run. Senserva scans the Microsoft AI surface in your tenant, every time, and turns it into ranked findings and audit-ready reports. That is the operational and performance evidence clauses 8 and 9 ask for.
For the detail on exactly what the scanner checks on the Microsoft AI surface, see Microsoft AI security.
Mapping Senserva to ISO/IEC 42001
Senserva does not replace your AIMS. It supplies the Microsoft-side evidence for the controls that are hardest to keep current.
| ISO/IEC 42001 area | How Senserva provides evidence |
|---|---|
| AI risk assessment (Clause 6) | Ranked findings across the Microsoft AI surface, prioritized by real-world risk, give you a repeatable input to the risk assessment. |
| AI system lifecycle and use (Annex A) | Continuous configuration checks on Copilot and AI agents show the controls around how AI is deployed and used are actually in place. |
| Data for AI (Annex A) | Identity, access, and SharePoint posture surface where an AI assistant could reach data it should not. |
| Third-party AI (Annex A) | App registrations, OAuth grants, and service principals for AI apps are audited for excessive permissions. |
| Performance evaluation (Clause 9) | Scheduled re-scans and audit-ready reports give you the monitoring and internal-audit evidence, with a clear before and after when you fix something. |
| Improvement (Clause 10) | Validated, approve-before-apply remediation closes findings, and the next scan proves the nonconformity is resolved. |
The AI you build should be governable too
ISO/IEC 42001 also asks that your own use of AI be responsible. Senserva's AI features are built that way on purpose: you bring your own model, your data stays on your machine, every answer is grounded in your real findings, and every remediation is reviewed before it is applied. That is the posture an assessor wants to see, and it is documented in Senserva Trustworthy AI.