EU AI Act enforcement begins August 2026. MaRisk, FCA, BaFin, and ACPR already require auditable AI decisioning. CB-12 packages the receipt chain, the drift monitor, and the governance documentation banks and insurers need to keep AI models in production without taking them down.
Every regulated AI deployment has the same problem: the model works, the regulator wants the paper trail, and the internal audit team wants a way to prove the model has not drifted since the last review. CB-12 plus the Spectral Drift Monitor close that loop without rewriting the model.
Credit decisioning, fraud scoring, transaction monitoring. MaRisk-aligned model risk management. BaFin-ready receipt chain for every automated decision.
Auto, life, health, commercial. Article 26 obligations met for high-risk AI. Forensic replay for disputed claims and adverse-action notices.
High-risk AI under EU AI Act Annex III. Decision explainability with hash-chained receipts. Adverse-action evidence the regulator and the borrower both accept.
Receipt-bound clinical decisioning. Drift monitoring tied to model registration. Audit packets for hospital compliance and notified-body review.
Resume screening, candidate ranking, performance scoring. High-risk under Annex III. Bias-monitoring evidence packets pre-built for works-council review.
FCA, BaFin, ACPR-aligned. Receipt chain for every alert, suppression, and escalation. Audit-grade evidence the financial supervisor can ingest.
Personalization and dynamic pricing under DSA, GDPR, and consumer-protection law. Receipt-bound rationale for every personalized decision.
MiFID II suitability evidence. Receipt-bound investment recommendations. Forensic replay for client disputes and regulator examinations.
Read-only Federal Audit Console skin for second-line and third-line teams. Continuous monitoring instead of point-in-time validation.
Article 12 (logging), 13 (transparency), 14 (human oversight), 26 (deployer obligations) packaged as a deployable appliance. 77 tests passing. SKUs at $25K Pro, $75K Enterprise (5M receipts), $250K Enterprise (50M receipts), on-prem Sovereign on quote.
Drift detection plus integrity attestation. Wraps any model endpoint, emits hash-chained receipts, and surfaces spectral-coherence anomalies before a regulator-visible incident.
whl-governance v0.2.0. 696 tests passing. Drop-in proposal gate, evidence bundle schema, and receipt ledger client. PyPI-ready and embeddable in existing model-serving pipelines.
Hosted receipt-chain endpoint for teams that do not want to operate the appliance. Per-decision pricing. Audit packet export on demand for regulator response.
We run paid PoCs with banks, insurers, and high-risk AI deployers. Four to eight weeks from kickoff to a working CB-12 appliance against one of your production models. Compliance, model-risk, and procurement all loop in.