ISO 27090 vs ISO 42001 explained
Jonathan BezdekCTO
5 Min. read
ISO 27090 guides AI security work. ISO 42001 sets requirements for an artificial intelligence management system, or AIMS. Organisations can use them together to connect security practices with wider AI governance. Sources: ISO 27090 publication record, ISO 42001 publication record.
ISO 27090 vs ISO 42001 at a glance
| Comparison | ISO 27090 | ISO 42001 |
|---|---|---|
| Main purpose | Protect AI systems against security threats | Organise the responsible development and use of AI |
| Type of standard | Security guidance | AI management system requirements |
| Practical question | How could this AI system be compromised? | How do we govern AI consistently across the organisation? |
| Main focus | Threats to AI data, models and workflows | AI policies, responsibilities, risks, objectives and improvement |
| Example concerns | Data poisoning, model theft and related attacks | Security, safety, transparency, bias and other AI risks |
| Certification | Guidance; does not establish an AIMS certification route | Voluntary certification of the organisation’s AIMS |
| Who can use it | Organisations developing or using AI | Organisations developing, providing or using AI |
| Current edition | ISO/IEC 27090:2026 | ISO/IEC 42001:2023 |
| First published | 14 October 2026 | 18 December 2023 |
Sources: ISO 27090 publication record, ISO 42001 publication record and ISO explanation of ISO 42001.
Their scope overlaps. ISO 42001 includes AI security among broader organisational concerns. ISO 27090 also addresses governance in the context of AI security. Their different purposes explain how they can complement each other. Sources: ISO explanation of ISO 42001, ISO explanation of ISO 27090.
ISO 27090 and ISO 42001 certification
ISO 42001 certification assesses whether an organisation’s AI management system meets the standard’s requirements. Independent certification bodies perform certification; ISO itself does not certify organisations. Individual AI applications still need security assessment. Source: ISO explanation of ISO 42001.
When commissioning ISO 27090 work, define the security review you need: which systems, which attack scenarios and what evidence will support the findings. This gives the assessment a concrete scope and a useful outcome.
Which standard should you use
Choose according to the problem you need to solve. This table is a practical decision guide.
| Your situation | Where to focus | Recommended next step |
|---|---|---|
| AI use has no clear owners or rules | ISO 42001 | Establish accountability, policy and a review process |
| A model or application faces specific attack risks | ISO 27090 | Analyse attack scenarios and select safeguards to test |
| A customer requests AI management system certification | ISO 42001 | Define the certification scope and assess readiness |
| AI governance exists, but security evidence is weak | Add ISO 27090 guidance | Connect security risks to controls and test results |
| You are launching an AI service and building governance | Use both together | Coordinate management responsibilities and security work in one programme |
An urgent application security issue deserves immediate attention. A broader governance programme can develop alongside the work needed to address that exposure.
How to use ISO 27090 and ISO 42001 together
For example, a company introducing an AI support assistant could coordinate its work as follows. These are illustrative implementation choices.
| Activity | How ISO 42001 can structure the work | How ISO 27090 can inform security work |
|---|---|---|
| Plan the service | Assign an owner and agree its purpose | Identify sensitive assets and attack paths |
| Review suppliers | Record responsibilities and review criteria | Examine exposure through models, data and dependencies |
| Approve release | Record risk decisions and the responsible approver | Test relevant attacks and application safeguards |
| Manage changes | Review outcomes and decide improvements | Reassess security when models, data or integrations change |
Keep one AI system record with links to its owners, risks, controls, tests and decisions. Reuse evidence where it is relevant, and record which assessment criteria it supports.
For example, a security test can show whether a restricted action was blocked. The associated review record can show who evaluated that result and decided whether the system was ready for release. Both records are useful, and they answer different questions.
For how ISO 42001 relates to the NIST AI RMF and the EU AI Act, see ISO 42001, NIST AI RMF, and the AI Act: connect the roles.
Frequently asked questions
Does ISO 27090 replace ISO 42001?
No. They serve different purposes: AI security guidance and AI management system requirements.
Can organisations using third-party AI apply them?
Yes. Both standards address organisations that use AI.
Does ISO 42001 certification prove every AI application is secure?
No. It assesses the AIMS. Application security still needs evaluation.
Sources: ISO explanation of ISO 27090 and ISO explanation of ISO 42001.
Connect AI governance with security evidence
Start with the outcome you need: clearer AI accountability, stronger application security or a combination of the two. Give the work an owner and decide what evidence will demonstrate progress.
Secani connects requirements, risks, controls, evidence, owners and review decisions in one workspace. Talk to Secani about organising your AI governance and security work.
Build auditable compliance workflows
Secani connects scopes, evidence, tasks, and AI agents in one shared workspace.
Related posts
All postsAI governance
ISO 27090 explained for AI security teams
A practical guide to the new AI cybersecurity standard, the attacks it addresses and the evidence security teams can use.
AI governance
ISO 42001, NIST AI RMF, and the AI Act: connect the roles
Use AI management systems and risk frameworks to organize governance while assessing the AI Act’s legal requirements separately.
AI governance
EU AI Act: classify your use cases and plan the evidence
A current, role-based approach to AI Act readiness, including the Commission’s updated timetable, high-risk assessment, and accountable governance.