Degradation of an AI model's performance over time as real-world conditions change from what it was trained on. Like skills becoming outdated as the world changes.
A fraud detection model experiences drift as scammers develop new tactics that weren't in the training data, requiring periodic retraining.