Model drift is the gradual decline in a machine learning model’s accuracy over time, as the real-world data it sees in production drifts away from the data it was trained on.
In production, the world keeps moving. Customer behavior, fraud patterns, prices, and language all shift while the model stays frozen at its training snapshot. Predictions that were accurate at launch slowly degrade, often with no error and no alert, until a business metric starts slipping and someone goes looking for the cause.
Model drift is often a symptom rather than the root cause. A lot of what looks like the model “going stale” is really the input data quietly changing shape upstream. So the most durable response usually lives in the data layer rather than in retraining alone, which means keeping the data an AI uses usable, traceable, and reproducible. It is also distinct from execution drift, where the same model and data return different results because the run environment or data state changed.