The agent applies the definition established by the interagency model risk management guidance: a model is any quantitative method, system, or approach that applies statistical, economic, financial, or mathematical theories, techniques, and assumptions to process input data into quantitative estimates. Under this definition, the institution's model inventory includes credit scoring and underwriting models, fraud detection algorithms, AML transaction monitoring systems, interest rate risk and liquidity models, prepayment and credit loss estimation models, and the AI agents operating in the institution's compliance, operations, and customer-facing programs.
The explicit inclusion of AI agents in the governance scope is what distinguishes a current model risk program from one written for the pre-AI generation of quantitative tools. The April 2026 interagency guidance updated the principles to address AI and machine learning models specifically, including the additional validation challenges that complex AI models present, the explainability requirements for AI-assisted adverse actions, and the ongoing monitoring obligations for models whose behavior can drift in ways that simpler statistical models cannot. The agent's governance framework reflects these updated requirements.
