AI Glossary
Risk Tolerance
The acceptable level of variation or deviation an organization is willing to allow for a specific risk.
Risk Tolerance
Overview
Risk appetite and risk tolerance are closely related, but they are not the same.
Risk appetite describes an organization’s overall willingness to accept risk.
Risk tolerance defines the acceptable limits for specific risks.
A helpful way to think about risk tolerance is driving within a speed limit.
The destination remains the same, but there are clear boundaries that should not be exceeded.
Organizations establish similar boundaries for AI systems.
For example, they may define acceptable thresholds for model accuracy, fairness metrics, response times, security incidents, or compliance violations.
If an AI system exceeds those limits, corrective action may be required.
Clearly defined risk tolerances help organizations monitor AI systems consistently and respond before issues become significant.
Risk tolerance is an important part of effective AI governance and operational oversight.
Why It Matters
Risk tolerance establishes measurable limits that help organizations manage AI systems consistently.
Real-World Example
An organization may require an AI fraud detection model to maintain at least 98% accuracy before additional review is triggered.