Which is an AI-specific step in incident response?

• A. Data collection for network logs
• B. Hardware failover testing
• C. Drift detection
• D. Routine backups

Answer: (C)

In AI incident response, keeping an eye on model drift is essential. Drift detection monitors changes in the data the model sees and the relationship between inputs and outputs. If the incoming data distribution shifts (data drift) or the relationship the model learned (the concept) changes (concept drift), the model’s predictions can become unreliable or reveal data integrity or security issues. Detecting this early allows the incident response team to investigate data pipelines, potential data poisoning, or adversarial manipulation, and to retrain or adjust the model accordingly.

Data collection for network logs, hardware failover testing, and routine backups are general IT and incident-response practices that apply regardless of AI, not something unique to AI systems. Drift detection, by contrast, is specific to monitoring AI/ML systems and their performance in the face of changing data, making it the AI-focused step in incident response.