What phenomenon occurs when AI models trained on historical data become less accurate over time due to new data?

• A. Concept drift
• B. Data lag
• C. Model drift
• D. Data leakage

Answer: (C)

Concept drift describes what happens when the relationship between inputs and outputs changes over time, so a model trained on historical data becomes less accurate as new data come in. As patterns in the environment evolve—like shifting consumer behavior, new fraud tactics, or changing sensor distributions—the model’s learned associations no longer match the current data, causing performance to drop. You might notice declines in accuracy, changes in calibration, or shifts in which features matter most. Data leakage is about training with information that wouldn’t be available in production and isn’t about performance decay over time. Data lag isn’t a standard term for this phenomenon. Model drift is sometimes used in different contexts, but the established term for degradation due to changing data distributions is concept drift. To handle it, retrain on recent data, update features, or implement drift detection and adaptive learning methods.