Concept Drift is when the relationship between a model’s inputs and the target it predicts changes over time. The model code stays the same, but what the inputs mean for the outcome shifts, so a model trained on past patterns gradually loses accuracy.
It differs from data drift, which is a change in the input data’s distribution rather than in the input-to-target relationship. Both are easier to catch when each run is bound to a fixed, released data state you can compare against.