OSFI E-23
Model Risk Management
Model risk is our first-class concern. The decision layer is deterministic and traceable; language-model outputs are confined to narrow, auditable roles and never produce portfolio weights.
Every model in use has a validation file, a monitoring plan, and a retirement date. Dissents are logged, not argued. Challenge is structural, not ad-hoc.