Glossary
Model risk
The risk of loss from decisions based on a model that is wrong, or that is used incorrectly or outside its design.
Model risk is the chance of adverse results from relying on a model. The US Federal Reserve guidance SR 11-7 (2011, replaced in 2026 by SR 26-2) names two main sources: fundamental errors that give inaccurate outputs, and use of the model incorrectly or outside what it was built for.
It is managed by independent validation, monitoring after launch and clear ownership of each model.
Example: a rule has a backtest daily mean of 0.05% and volatility of 1%. After 60 days it is down 9%, a z-score of -1.55 against that expectation. That is a warning to review, not yet proof of failure.