Accuracy disclosure
Infrarails issues per-deployment accuracy declarations rather than a single platform-wide number. A single number across all customers, all evaluators, and all benchmarks is not a defensible claim — real accuracy depends on which evaluators are enabled, which thresholds are configured, and what data the system is observing.
What you receive — a signed accuracy declaration scoped to your organisation, your evaluator configuration, and the benchmark suites you elect. Updated on every benchmark refresh.
When you receive it — during onboarding, as part of your EU AI Act Article 15 / NIST AI RMF MEASURE compliance package, and on request before each annual audit.
How it differs from a marketing claim — every disclosure ties to a reproducible benchmark run with a content hash. You can replay it.