Model Drift
Model drift is the degradation of AI model performance over time, occurring when real-world data, user behavior, or usage patterns diverge from the distribution the model was trained on.
How Model Drift Happens
An AI model captures a snapshot of the world as represented in its training data. The world moves on. Customer behavior shifts, new terminology appears, edge cases evolve, upstream data sources change. The model's outputs, which were accurate on training data, become progressively less accurate on current data.
Two related patterns are typically discussed: data drift (the distribution of inputs the model sees in production changes) and concept drift (the underlying relationship between inputs and correct outputs changes). Both degrade performance, and both require monitoring to detect.
Why Model Drift Is a Governance Concern
A model that passed pre-deployment validation and produced accurate results at launch is not guaranteed to still be performing well six months later. Governance programs that assume "approved once, approved forever" miss drift entirely. The NIST AI RMF treats ongoing performance monitoring as a core operational requirement, and the EU AI Act requires post-market monitoring for high-risk systems.
Detecting and Managing Model Drift
Effective drift management typically combines continuous AI monitoring of input distributions, output distributions, and performance metrics; scheduled reassessment against fresh validation data; and defined retraining or model-replacement procedures when drift crosses thresholds.
Related Terms
Full AI Visibility. Full Control. One Connected Platform.
Enterprise AI is expanding faster than most governance programs can track. Kovrr connects every AI signal across browser, endpoint, network, identity, and vendor systems into a single platform so security, governance, and risk teams work from the same evidence.
