Audit-grade artefacts, not templates.
Annex IV technical files, FRIAs, DPIAs, model cards and risk-management files generated from the registry — with a citation behind every claim, and an independent review before the document ever reaches you.
Template-shaped Word files do not survive an audit.
Most AI governance documents in circulation are templates with the right headings and the wrong contents — claims about systems no one verified, dates that pre-date the model in question, references to data the team no longer uses. They pass a checklist tool. They fail a regulator.
An artefact is only audit-grade if every claim in it can be traced to a fact in the system of record, and if the document is held open rather than padded when a fact is missing.
Sourced, reviewed, blocked when missing.
Every artefact is generated against the registry record, line by line, with the citation visible. An independent review runs before you ever read the draft.
Documents are a view of the registry, rendered to the format the framework demands.
The registry is the source. Documents render that source into Annex IV files, FRIAs, DPIAs and model cards — so the artefact, the classification, the questionnaire answer and the trust page entry all say the same thing because they all derive from the same record.
Sourced line by line. Reviewed before you see it. Blocked when a fact is missing.
A document is either traceable to the registry or it is not in scope. The system declines to invent. That is the difference between an artefact a regulator will accept and a template a vendor sold you.
See an Annex IV file generated against a real system.
Bring one live AI system. We'll register it on the call, classify it, and generate the artefact pack — with every claim cited and the missing-input rule visible.