Model Drift

The degradation of a model's predictive performance over time due to changes in the underlying data distribution or relationships between features and the target variable.

When an AI's performance slowly gets worse over time because the real world has changed. A fraud detection AI trained on 2020 data may miss new types of fraud that emerged in 2024.

Model drift is a key risk that governance frameworks must address. Without ongoing monitoring, AI systems can silently degrade, making incorrect decisions. Your governance framework should mandate drift detection and define retraining triggers.