A financial organization implements a new AI-based fraud detection system to flag suspicious transactions.

Basic Concept: When an AI fraud detection model produces false positives (blocking legitimate transactions), this indicates the model ' s decision boundary is insufficiently calibrated. The model needs improved training data to better distinguish fraudulent from legitimate transactions. CompTIA SecAI+ covers model performance improvement under AI-assisted security.

Why A is Correct: Retraining the model with more data and recent transaction patterns directly addresses the root cause of false positives. Additional representative legitimate transaction data helps the model learn more accurate decision boundaries, reducing false positives while maintaining detection sensitivity for actual fraud. This improves model accuracy without abandoning AI-based detection.

Why B is Wrong: Token usage and rate limits are cost and resource management controls for LLM APIs. They have no relevance to improving the accuracy of a fraud detection ML model that incorrectly classifies legitimate transactions.

Why C is Wrong: Encrypting data and applying access controls are data security measures that protect confidentiality and integrity. They do not address model classification accuracy or the false positive problem in fraud detection.

Why D is Wrong: Rolling back to a traditional system abandons the capabilities of AI-based fraud detection. The appropriate response to model performance issues is to improve the model through retraining, not to regress to less capable detection approaches.