What does ISO 42001 certification require for AI security?

ISO/IEC 42001:2023 requires an AI management system: a defined scope, an AI policy signed off by leadership, assigned roles, a documented risk assessment and treatment process, an AI system impact assessment, applicable Annex A controls with a statement of applicability, internal audit, and management review. Certification comes from an accredited body through a Stage 1 documentation audit followed by a Stage 2 implementation audit, then surveillance across a three-year cycle.

Security-relevant obligations concentrate in a few places. The AI risk assessment must address AI-specific risks rather than reusing your information security register, which means adversarial manipulation, model behavior, data provenance, and autonomy belong in scope. The impact assessment must consider effects on individuals and groups, not just the organization, which is where it diverges most sharply from ISO 27001 habits.

Annex A controls cover the AI lifecycle: data management and quality, system design and development, verification and validation, deployment and operation, and third-party and supplier relationships. The supplier controls matter more than teams expect, because most enterprise AI is someone else's model reached through someone else's API.

The requirement that generates the most remediation work is evidence of operation. Auditors want records showing controls ran, which means dated evaluation results, logged runtime enforcement, and change records, not a policy document describing intent.

If you hold ISO 27001, the harmonized structure means clauses 4 through 10 are familiar and Annex A is the new work.

Akto Argus and Akto Atlas generate the operational evidence these clauses require, including continuous testing results and runtime enforcement logs for AI systems and employee AI usage.

The certificate proves the system exists. Evidence proves it ran.