False negative

Spec sheet // Glossary // FALSE NEGATIVE
Glossary / F / False negative
Glossary · Transaction monitoring

False negative

Also called missed alert, type II error

Definition

A false negative is a real risk that a screening or monitoring system fails to flag. In sanctions screening, it is a true match to a listed person that produces no alert. In transaction monitoring, it is suspicious activity that passes unreviewed. False negatives let prohibited or criminal activity through.

REAL RISKSCREENINGNO ALERTMISSED
Fig. 1 · False negative
NYDFS Part 504 · OFAC compliance framework
Key rulesNYDFS Part 504 · OFAC compliance framework
Applies toBanks, money transmitters and other institutions regulated by NYDFS
RegionsGlobal · US
01

How it works

Every detection system trades off false positives against false negatives. Raising thresholds or tightening fuzzy matching cuts alert volumes but lets more real cases slip through unseen. Because missed cases leave no alert, firms cannot count false negatives directly from production data. They estimate them by testing the system with known cases and by reviewing activity that was later found to be suspicious.

Common tests include seeding screening tools with sanctioned names and spelling variants to confirm they alert, running below-the-line testing on transactions just under each threshold, and looking back at confirmed suspicious activity reports, law enforcement requests and adverse media to see whether monitoring caught them. Results feed threshold tuning and model validation, and documented testing is what supervisors ask for when they assess a programme.

Step 1Seed known casesInsert test names and typologies that should alert.
Step 2Test below the lineSample activity just under thresholds for missed risk.
Step 3Look backCheck whether confirmed cases were flagged earlier.
Step 4MeasureEstimate the miss rate for each rule or list.
Step 5TuneAdjust thresholds and matching, then retest.
02

The rules

US · Effective 1 January 2017; first certification 15 April 2018NYDFS Part 5043 NYCRR Part 504, Banking Division Transaction Monitoring and Filtering Program Requirements and Certifications

Requires transaction monitoring and watchlist filtering programmes with pre- and post-implementation testing of data, models, scenarios and matching, and an annual compliance certification.

Banks, money transmitters and other institutions regulated by NYDFS
US · 2 May 2019OFAC compliance frameworkA Framework for OFAC Compliance Commitments, Office of Foreign Assets Control

Names testing and auditing as one of five essential components of a sanctions compliance programme, with independent testing to find weaknesses and prompt fixes.

US persons and others subject to US sanctions
03

Common questions

What is the difference between a false positive and a false negative?

A false positive is an alert on something that turns out to be harmless. A false negative is a real risk that produced no alert. False positives waste analyst time; false negatives let real risk through.

How do firms measure false negatives?

They cannot count them directly, since missed cases leave no alert. Firms estimate them by seeding known cases, sampling activity below thresholds and checking whether later confirmed cases were caught.

Can AI reduce false negatives?

Machine learning models can find patterns that fixed rules miss, which can lower false negatives. They still need validation and ongoing testing to show they catch what they should.