Data drift occurs when the statistical properties of input data change over time, causing the AI model to produce increasingly inaccurate or biased results.
The correct response is to document the risk and implement continuous monitoring (A) because:
* Drift requires ongoing detection , not a one-time fix
* Retraining too early can reinforce issues (especially with the same dataset)
* Disabling the system (D) is overly disruptive unless drift results in unsafe decisions
* Continuing deployment without monitoring (C) increases risk of degraded performance AAIA emphasizes drift monitoring dashboards, alerting mechanisms, and retraining triggers as essential controls for AI operations.
References:
AAIA Domain 2: AI Operations - Monitoring, Drift Detection, and Lifecycle Maintenance