What is Model Risk Management?

Model Risk Management (MRM) is a discipline focused on identifying and mitigating risks associated with using quantitative models in business decisions. Originally developed in financial services (SR 11-7), it's increasingly applied to AI and machine learning models.

Sources of Model Risk

Model Error

Incorrect assumptions
Mathematical errors
Data quality issues
Implementation bugs

Misuse

Using models outside intended scope
Misinterpreting outputs
Over-reliance on models

External Changes

Market regime changes
Data drift
Concept drift

MRM Framework

Model Inventory

Catalog all models
Classify by risk tier
Track model lineage

Model Development

Documentation standards
Development controls
Testing requirements

Model Validation

Independent review
Conceptual soundness
Outcome analysis
Benchmarking

Ongoing Monitoring

Performance tracking
Drift detection
Periodic review
Change management

Regulatory Guidance

Financial Services

SR 11-7 (Federal Reserve)
OCC 2011-12
Basel requirements

AI-Specific

EU AI Act
NIST AI RMF
Industry guidelines

AI/ML Considerations

Explainability requirements
Bias and fairness testing
Continuous monitoring
Version control
Reproducibility