Model risk management checklist
Model risk work turns AI uncertainty into a repeatable control process: inventory, validation, monitoring, issue handling, and accountable signoff.
Inventory
Track model purpose, version, owner, vendor, data sources, deployment surface, and user impact.
Validation
Evaluate performance, robustness, bias, privacy, security, and misuse risk against the intended use.
Controls
Define human review, access limits, fallback behavior, incident severity, and change management.