What does an AI adverse-action claim cost your firm?
Six inputs. Two minutes. An IFRS 9 / ASC 450-defensible estimate of your §1002.9 reserve exposure — and what an independent decision record removes from it.
Lending
HR / AEDT
Healthcare
AI decisions per year500,000
1K5M
Adverse-action rate18%
1%60%
Primary enforcement jurisdiction1.5×
⚠ NJ: AG, DCR Director, and private complainants all have standing — the broadest enforcement surface of any US jurisdiction.
Evidence maturityLogs only
No evidenceFull DPR
Months since last model validation14 mo
Just tested3 years ago
Your modeled reserve exposure
Modeled §1002.9 reserve exposure — before GateFrame
—
Reserve after an independent decision record
—
GateFrame removes adverse-action notice accuracy as a failure mode — the specific Earnest violation.
Reserve reduction
—
GateFrame ACV
$80K
Reduction per $1 ACV
—
ASC 450-20 / IFRS 9 contingent-liability basis
Severity is a probability-adjusted expected cost per adverse-action notice — a blended value across contested and uncontested notices, anchored to the Earnest settlement over its affected portfolio. This models the §1002.9 notice-accuracy failure mode only; disparate-impact, proxy-discrimination, and training-data-lineage claims are out of scope.
Anchor case — Earnest Operations LLC: $2.5M settlement (MA AG, July 10 2025). Violation: inaccurate AI adverse-action notices that failed to state specific reasons. The settlement requires an internal algorithmic-oversight regime for fair-lending testing and accurate adverse-action notices. EU AI Act — Art. 12 (high-risk AI): record-keeping and traceability required.
Coverage scope
✓ Covered: §1002.9 notice accuracy · Earnest remediation · EU AI Act Art. 12 · CFPB Form C-1 ✗ Not covered: disparate-impact testing · training-data lineage · model-performance monitoring Pair with your ML platform's bias auditor for full §1002 coverage.