One page that answers the board question: "what do we actually know about the AI we deploy?" The record below is measured, signed, and independently verifiable — and it stays measured, never certified.
Loading live board…
| For the board | For procurement | For the 2028 Illinois audit |
|---|---|---|
| Which provisions our AI systems are measured against | Deterministic conformance scores with 95% CIs | A signed, gapless chain an auditor can walk offline |
| Who leads each axis (no LLM-as-judge) | Sample counts per bank (237, 36, 32…) | Reproducible metrics — re-run and compare |
| Market context reported alongside, never fused | Corrections ledger (appended, never edited) | Version history + timestamps |
| Date | Obligation | Penalty exposure |
|---|---|---|
| 11 Sep 2026 | EU CRA Art 14 reporting | €15M / 2.5% turnover |
| 1 Jan 2027 | Illinois SB 315 disclosure · NY RAISE · Colorado SB 26-189 | $1M first / $3M subsequent (IL) |
| 1 Jan 2028 | Illinois SB 315 independent third-party audit | $1M / $3M |
Fetch a card → get the trust root (csoai.org/.well-known/did.json) → check the Ed25519 signature → check the hash chain → re-run a metric. Or use the front door verify box.
meok.ai/cards/h3k-2026-08-20T0422.json · methodology DOI 10.5281/zenodo.21991104