Every dataset ships with an Ed25519-signed cert, a 19-dim LQS quality report, contamination-clean flags against 40+ public benchmarks, and a revocation ID your CI can poll. Built for SR 11-7, EU AI Act Art. 10, and 21 CFR 11 documentation your risk team already files.
Price includes the raw data, the signed cert, the quality report, and the registry entry. Nothing you pay extra for. Nothing that requires a call.
Cryptographically signed certificate containing the 19-dim quality breakdown, provenance chain, license text, and revocation ID. Verifiable offline against our public key aa4c070af907e2ea.
Structural integrity, annotation quality, statistical health, training fitness, provenance, subgroup equity, oracle agreement — each with a per-dim confidence interval you can drop into a model validation package.
Every dataset is cross-checked against 40+ public evaluation sets. The cert lists per-benchmark contamination flags, so you know what you can train on before you burn compute on a leaked eval.
Cert revoked post-release because of newly-discovered contamination or a provenance dispute? Your CI polls the public registry by cert ID. Failure modes discovered a year from now don't require a support ticket.
Perpetual commercial license, attached to the receipt. No "check each dataset's README" workflow. Legal can sign off before you schedule training time.
pip install labelsets. One-line loader, cert-verify step in CI, GitHub Action for release gates. Same ergonomics as datasets.
Every cert fits the shape of a standard vendor-evidence line item. Copy/paste into your SR 11-7 validation template or EU AI Act Art. 10 data-governance filing.
Three flagship datasets live: legal reasoning ($799), financial routing ($549), clinical reasoning ($699). Every cert verifies against our public key. Perpetual license, no subscription, no call required.