Compliance is not a checkbox. It is a measurement, on a published harness, signed so a
third party can verify. This page is the operational guide for what we measure and how the
evidence is produced.
The four pillars of credible evidence
Frozen harness — the tests are published before the run, so the run cannot be tuned to pass.
Deterministic execution — pinned seeds, published splits, confidence intervals on every figure.
Signed artefacts — Ed25519-signed result records a third party can verify without trusting us.
Corpus-watch — when the provision text changes, the affected measurements re-fire. Compliance that goes stale is not compliance.
Article 43 (internal-control route) — most Annex III systems can self-assess; what makes the assessment credible is evidence, not assertion.
What we do not do
Issue certifications. We measure; a regulator certifies.
Replace notified bodies where the law requires them.
Sign "passed" claims without a measurement and a harness.
Issue accreditations — we are a measurement body, not an accreditation body.
Boundary
This is not legal advice. The evidence layer is verifiable and inexpensive; the legal
interpretation is its own layer (notified bodies, regulators, counsel). Get a lawyer for
anything that touches liability.