Detecting and Correcting Temporal Drift in Credit-Risk and Fraud-Detection Models at Community Banks, Credit Unions, and CDFIs: An Automated Model-Risk-Monitoring Framework Below the Supervisory Threshold

Abstract

On April 17, 2026, the Federal Reserve, OCC, and FDIC jointly issued SR 26-2, superseding SR 11-7 and introducing a uniform 30 billion U.S. dollar total-asset threshold across all three agencies, replacing the FDIC's prior 1 billion U.S. dollar threshold and the absence of any explicit threshold at the Federal Reserve and OCC [1][2]. This article proposes an automated model-risk-monitoring framework for detecting and correcting temporal drift in credit-risk and fraud-detection models at community banks, credit unions, and CDFIs, institutions now explicitly positioned below this new supervisory threshold, drawing on real, current evidence that 56 percent of financial institutions experienced significant model drift in at least one production model as of 2023 [4], and on published, peer-reviewed unsupervised drift-detection methods (D3 and OCDD) whose labelling-efficiency profiles, 10 versus 75 samples per detected drift respectively, directly determine feasibility for smaller, resource-constrained institutions [3]. The article details the automated framework, its alignment with SR 26-2's new materiality-based, size-scaled approach to model risk management, and the evidence limitations that remain for institutions below the new threshold specifically.