Why individual guardrails alone aren't a structure
"We've built in guardrails" has become a standard line in almost every conversation about AI projects – usually meaning individual technical tricks: a prompt filter here, an output check there. The problem isn't that these measures are wrong; it's that without an overarching structure, they're hard to communicate, hard to audit, and hard to demonstrate in a customer or investor conversation.
The two frameworks in short
The NIST AI RMF is a free, voluntary guide from the US standards body NIST, built on four functions: Govern (organization, responsibilities, approval processes), Map (understanding the context and possible harms of a specific AI system), Measure (making risks measurable with metrics and tests), and Manage (deploying resources, prioritized, to address measured risks). It isn't certifiable – you don't "pass" it, you apply it.
ISO/IEC 42001, by contrast, is a genuine certification standard – the first international standard for an AI management system. It follows the same structure as ISO 27001 (information security) and ISO 9001 (quality): a company that already runs one of those systems builds an AI management system considerably more easily than one starting from zero.
The myth: ISO 42001 as legal protection
The EU AI Act's Article 40 provides a "presumption of conformity": companies working to a recognized standard may assume certain obligations are met. The catch is in the word "recognized" – legally, that only applies to a standard that has gone through the full European harmonization process. No AI standard has that status yet, including ISO 42001 (see "The EU AI Act – what actually affects your business"). An ISO 42001 certification is therefore – as of today – not automatic legal protection.
The real gatekeeper: procurement, not the legislator
Companies increasingly buy AI use cases ready-made instead of building them themselves. That's exactly what's reshaping vendor questionnaires: regulated large enterprises now explicitly distinguish between "we have an AI policy" (weak, see "Do we need an internal AI policy?") and "we run an AI management system" – precisely the structure NIST AI RMF and ISO 42001 formalize. A company that can't answer that structure doesn't primarily risk a fine – it loses the bid.
From function to guardrail – the translation table
The four NIST functions are deliberately abstract so they fit any type of AI system – valuable as a communication framework, but useless as a build guide. For a concrete project, each function translates into a category of technical and organizational guardrails: Govern becomes organizational guardrails (approval processes, responsibilities), Map becomes context guardrails (which data, what possible harm per feature), Measure becomes technical guardrails (output validation, tests, monitoring), and Manage becomes response guardrails (rollback capability, escalation path, review cadence).
Why this matters for you as a decision-maker
For most companies, the practice matters before the certificate: the Govern/Map/Measure/Manage structure can be applied within an ongoing project without waiting for an audit – and it's exactly what a vendor questionnaire wants to see. The full ISO 42001 certificate only becomes relevant once a customer explicitly requires it. The full picture, market data, and a more detailed version of the translation table are in the article "NIST AI RMF and ISO 42001: what AI governance frameworks actually deliver – and what they don't (yet)".