Security questionnaires now ask a new set of questions: which AI systems do you use, what data do they see, who is accountable for their outputs, how do you assess their risks and how would you know if one went wrong. Regulators in several jurisdictions are writing obligations for providers and deployers of AI systems, and buyers are asking for evidence before the law requires it. ISO/IEC 42001:2023, the international standard for an AI management system, is the structure most organisations are using to answer.
It is a management system, and that is the point
ISO/IEC 42001 follows the harmonised structure shared by ISO 9001 and ISO/IEC 27001: context, leadership, planning, support, operation, performance evaluation and improvement, applied to the organisation's development, provision or use of AI. It adds what AI specifically needs: an AI policy and objectives, defined roles and accountability for AI systems, an AI risk assessment and an AI system impact assessment that considers individuals, groups and society, controls over the AI system life cycle and over data, and transparency to the people affected. An organisation that already holds ISO/IEC 27001 will recognise the machinery and can integrate the two.
You do not have to build AI to need it
The standard applies to organisations that use AI as much as to those that develop it. A professional services firm using generative tools on client material, a retailer using recommendation and pricing models, a recruiter using screening software, and a software company embedding third-party models in its product all make decisions with AI that affect people, and all are being asked how those decisions are governed. Scope the system to the AI you actually use and the roles you actually play (provider, deployer, or both), and the standard becomes proportionate.
Five first steps
Inventory: list every AI system in use or development, its purpose, its data, its supplier and its owner. Most organisations find more than they expected, and the list is the foundation of everything else.
Roles: decide who is accountable for AI at board level and who owns each system; write it down.
Risk and impact: for each system, assess the risks to the organisation and the impact on the people affected, in proportion to what the system decides; the high-impact systems get the attention.
Controls and life cycle: define how systems are specified, tested, approved, monitored and retired, and how data is governed, using the standard's reference controls as the menu.
Transparency and incidents: decide what people are told and how problems are reported, investigated and corrected.
What certification proves
Certification by an independent, accredited certification body tells a buyer or regulator that the AI management system exists, is operated and is audited, in the same way an ISO/IEC 27001 certificate does for security. It does not certify that a particular model is accurate or fair; it certifies that the organisation governs its AI in a way that would detect and correct a problem. That is the assurance procurement teams are asking for, and it is the assurance an AI-specific law will expect to see behind a conformity claim.
What to do next
Start with the inventory and the accountability decision; both cost nothing and both are asked about first. Download the ISO/IEC 42001 readiness checklist for the rest, and read the ISO/IEC 42001 certification page for scope, timeline and how it integrates with an existing ISO/IEC 27001 certificate.
Certification with TRAIBCERT: ISO/IEC 42001
AI management system. Tell us your sites, headcount and the date you need it by; a sector specialist replies within one working day with a fixed quotation for the whole cycle.
