Thought leadership
The False Negative Problem
What the alert reduction headlines are not telling you
- Topic
- Transaction monitoring and model performance
- From
- Coalition Against Financial Crime
- Length
- 22 pages

90 to 95%false positive rate that industry data places on rules-based monitoring systems
1 in 20alerts an investigator reviews that is likely to reflect genuine suspicious activity
70 to 90%+alert volume reductions promised by machine learning and agentic AI vendors
?false negatives: the half of the performance equation with almost no structured attention
The argument
The scrutiny
Not because false positive reduction is unimportant, but because the industry conversation has concentrated so heavily on one half of the performance equation that the other half has received almost no structured attention.
No trace
False negatives leave no trace by design.
Inside this paper
- Introduction: The Incomplete Story
- Understanding the Measurement Asymmetry
- What Vendor Benchmarks Actually Measure
- The Agentic AI Inflection Point
- Questions Every Institution Should Be Asking
- The Standard That Should Govern This Space
- Closing the Knowledge Gap: Certified AI Practitioner in Financial Crime (CAPFC)
- Conclusion
Get your free copy
Enter your details and we will email you the white paper.
Coalition Against Financial Crime
An independent professional organization advancing the fight against financial crime through collaboration, practical education, innovation, and action.
cafc.org
An independent professional organization advancing the fight against financial crime through collaboration, practical education, innovation, and action.
cafc.org
