In AI-enabled security monitoring, PMI-style AI risk management highlights false positives and false negatives as key operational risks: false positives overwhelm analysts and create alert fatigue, while false negatives hide real threats. To mitigate these, guidance stresses continuous monitoring, feedback, and human–AI collaboration, not just algorithm choice. Establishing a continuous feedback loop with security teams (option D) means that security analysts review alerts, label them as true/false, and feed those labels back into the AI pipeline. This enables threshold tuning, recalibration, and retraining, incrementally reducing misclassification rates over time.

Option B (model combinations and trade-offs) can help at design time, but it does not by itself guarantee ongoing control of false positives/negatives once the system is deployed. Option A is too narrow and algorithm-specific and ignores the governance and lifecycle aspects. Option C addresses data security, which is important but unrelated to classification error rates. PMI-style AI operations (akin to MLOps) underline that closed-loop learning with real-world feedback is critical for safety, resilience, and performance. Hence, the action that directly addresses the risk of false positives and false negatives is to establish a continuous feedback loop with security.