How it works

Any detection system balances sensitivity against precision. Broad rules and loose fuzzy-matching thresholds catch more true risk but also flag more legitimate activity. A false positive arises when an alert is reviewed and closed because the activity is explained (for example, a name match with a different date of birth, or a large payment consistent with the customer's known business). The opposite error, a false negative, is a missed true risk and carries regulatory and legal exposure.

Requirements by jurisdiction

No regulator sets a target false-positive rate, but each expects detection to be risk-based and well governed.

• US: The FFIEC manual expects filtering criteria and thresholds to be reasonable, tailored to the institution's risk, documented, periodically reviewed and independently validated. OFAC enforcement actions have cited both failures: weak matching that missed listed parties, and alerts wrongly closed as false positives because of resourcing backlogs.

• EU: Under the AMLR, applicable from 10 July 2027, monitoring must be proportionate to risk, and AMLA guidance will shape supervisory expectations for monitoring quality.

• UK: The MLR require risk-based monitoring, and the FCA has reviewed firms' sanctions systems and controls, including screening calibration.

Common challenges

Poor data quality, missing secondary identifiers, generic thresholds applied to all customer segments, and rules that are never retired all inflate false positives. Tuning too aggressively risks suppressing genuine alerts, so changes need evidence and approval.

How teams handle it in practice

Teams measure alert outcomes by rule, segment customers so thresholds match expected behavior, enrich data to discount name matches automatically, and test proposed threshold changes against historical true positives before deploying them. FATF's 2021 report on new technologies notes that machine learning can reduce false positives and help identify complex cases, provided it is explainable and well governed. Compliance platforms like Corsa can pre-investigate alerts with AI agents so analysts spend time on likely true positives.