AI governance
AI assurance without the theatre
ISO/IEC 42001 is being sold as a badge. Treated properly, it's the most practical tool we have for governing systems that learn, drift and surprise us.
There's a growing market in AI governance theatre: principles on the website, an ethics committee that meets twice a year, and a certificate on the way. None of it tells you whether the model making credit decisions on Tuesday behaves like the one you assessed in March.
ISO/IEC 42001 is valuable precisely because it refuses to let you stop at principles. It asks for an AI system inventory, impact assessments tied to real use cases, defined roles across the lifecycle, and monitoring that would actually detect drift.
An AI policy without an inventory is a promise about systems you can't name.
Where to start
Start with the inventory, and be honest about shadow AI. Then pick your two highest-impact use cases and run them end-to-end through impact assessment, control selection and monitoring before you scale the approach. That's how we structure IIP-42001: build an evidence release, have it reviewed, correct it, validate it, and only then move on.
Assurance is assurance. The disciplines haven't changed; the systems have. Teams that already run a mature ISMS are closer to 42001 than they think, provided they treat it as an operating loop, not a document set.