Lesson 3 of 5
AI and model risk management: inventory, validation and monitoring
Supervisors in the United States, the United Kingdom and Canada expect AI models that fall under their model risk guidance to be governed like any other model: listed in an inventory, rated by risk, checked by people who did not build them, and monitored while they run. What AI changes is how hard some of these steps become.
Published 14 October 2026 · Listen to this lesson
In this lesson
- What a model is, and what model risk is
- Start with an inventory
- Rate each model by materiality and complexity
- Validation by people who did not build the model
- Monitoring models that run, and models that learn
- Who answers for it, until it is retired