Data drift occurs when the statistical properties of input data change over time, impacting the accuracy of an AI model. In this case, the model failed to adapt to recent demand trends, indicating that the data on which it was trained no longer reflects current behavior.
"Data drift leads to performance degradation in AI systems when real-world input data shifts away from training data patterns. Monitoring for drift is essential, particularly in dynamic environments such as retail." Although training diversity (A) and outlier detection (D) are relevant to accuracy, only B addresses the temporal misalignment between training data and real-time data. Overfitting (C) would more likely cause poor generalization in other contexts.
Reference: ISACA Advanced in AI Audit™ (AAIA™) Study Guide, Section: "AI Operations and Performance," Subsection: "Model Monitoring and Data Drift Detection"