The worst audit outcome is discovering gaps during the assessment. A coverage report is like a pre-flight checklist: it shows you what is instrumented and what is missing before your auditor arrives.
Every green cell is a procedure with verified anchors. Every red cell is a gap your auditor will flag. The report turns abstract compliance requirements into a concrete engineering task list.
Gaps found before the assessment are fixable. Gaps found during the assessment become findings. The difference between those two outcomes is a coverage report.
This is a static mockup. The live coverage report is generated by the SWT3 SDK demo or the Axiom platform.
Each uncovered procedure appears as a gap card. The card tells you what is missing, why it matters, and where to find instrumentation guidance.
No drift monitoring anchors found in the assessment window. Your auditor checks this under EU AI Act Art. 9(2)(b) and NIST AI RMF MEASURE 2.6.
No fairness testing anchors found. Your auditor checks this under EU AI Act Art. 10(2)(f) and NIST AI RMF MEASURE 2.5.
No DPIA anchors found. Your auditor checks this under EU AI Act Art. 27 and GDPR Art. 35.
2 anchors found, but none in the last 30 days. Your auditor may flag this as stale evidence.
| Framework | Total Procedures | After Client Wrapping | After Full SDK Integration |
|---|---|---|---|
| EU AI Act (Regulation 2024/1689) | 80 | ~5 covered | ~58 covered |
| NIST AI RMF (AI 100-1) | 80 | ~5 covered | ~58 covered |
| CMMC / NIST 800-53 | 30 | ~5 covered | ~22 covered |
| SR 11-7 (OCC 2011-12) | 19 | ~3 covered | ~15 covered |
Coverage depth depends on which integration patterns you adopt. Client wrapping alone covers inference witnessing. Full SDK integration adds RAG, tool, identity, access control, drift, fairness, and model metadata witnessing.
Every green cell in your coverage report corresponds to a checked row in your auditor's assessment checklist at /registry/checklist.html. Every red cell is a procedure your auditor will mark as a gap.
The coverage report and the checklist are two views of the same data. Yours shows what you have built. Theirs shows what they need to verify. Closing the gap between these two views before the assessment is the purpose of the coverage report.
See the assessment mapping guide for a full walkthrough of how procedures map to assessment objectives.
The SWT3 SDK demo generates a local HTML coverage report showing which procedures have been demonstrated and which need production data. Run the demo locally to see your starting position. Connect to the Axiom platform to track coverage across your full inference history over time.
See the SDK documentation for quickstart instructions and the before and after guide for integration patterns.