Model risk management (MRM) is a discipline focused on managing risk from models themselves.
By defining how models are developed, validated, approved, monitored, and retired, MRM reduces model errors, supports safe and consistent model use, strengthens decision-making, and ensures that models are fit for purpose throughout their lifecycles.
Traditionally, the scope of MRM is narrower and more technical than AI governance, focusing specifically on the models that power decision-making.
Focus areas
MRM addresses several key areas:
Development and documentation standards
Establish clear standards for model development, including documentation of purpose, design, theory, and logic.
Independent validation and challenge
Ensure models undergo rigorous independent validation to identify limitations and verify performance.
Performance monitoring and drift
Monitor model performance over time, detecting drift and degradation that may affect reliability.
Change control and remediation
Manage model changes through controlled processes, tracking findings and remediation actions.
Typical artifacts
MRM programs produce documentation that demonstrates sound practices:
Model inventories — Centralized registries of all models and their metadata
Model documentation — Comprehensive documentation of model design, development, and testing
Validation reports — Independent assessments of model performance and limitations
Findings — Issues identified during validation with remediation tracking
Monitoring evidence — Ongoing performance metrics and drift analysis
Using ValidMind for MRM
ValidMind supports MRM workflows through a unified platform:
flowchart TB
center((ValidMind<br>Model Risk Management<br>Platform))
inv((Model<br>Inventory))
docs((Documentation<br>Automation))
val((Validation<br>Workflows))
lod((Three Lines<br>of Defense))
mon((Ongoing<br>Monitoring))
life((Lifecycle))
center --- inv
center --- docs
center --- val
center --- lod
center --- mon
center --- life
%% styling
classDef centerNode fill:#ffffff,stroke:#de257e,stroke-width:2px,color:#de257e,text-align:center
class center centerNode
Model inventory
Track all models with customizable fields for materiality, tiering, and risk rating.
Documentation automation
Generate model documentation through configurable templates and automated testing.
Validation workflows
Configure independent validation processes with findings tracking.
Three lines of defense
Support role separation for developers, validators, and auditors.
Ongoing monitoring
Monitor model performance with alerts and drift detection.
Lifecycle
Govern model development, deployment, monitoring, and retirement stages.
Regulatory frameworks
ValidMind is designed to help organizations comply with MRM regulatory requirements:
SR 11-7 (United States)
The Supervisory Guidance on model risk management was issued by the Board of Governors of the Federal Reserve System and the Office of the Comptroller of the Currency. It provides comprehensive guidance on designing and maintaining a robust MRM framework.
The Model Risk Management Principles for Banks, issued by the Prudential Regulation Authority (PRA), set out comprehensive expectations for managing model risk in UK banks. The principles emphasize governance, strategic planning, model development, validation, and robust technical capabilities.
The Office of the Superintendent of Financial Institutions (OSFI) issued the Guideline on Enterprise-Wide Model Risk Management (E-23) for Deposit-Taking Institutions in Canada. E-23 sets out prudent practices for model development, validation, approval, use, and ongoing change management, including considerations for AI and machine learning models.