The REGTECH Book

Can AI Really Disrupt Monitoring for Suspicious Activity?

By Malcolm Wright

Chief Compliance Officer, Diginex

At this moment, tens of thousands of anti–money laundering (AML) analysts are sitting glued to their screens poring through mountains of alerts generated by their transaction monitoring systems. Depressingly, most of these alerts will be false positives. Transaction monitoring in its current form is primitive, using a series of rules to detect suspicious activity. Such rules might include monitoring for a cash payment or series of structured payments that exceed US$10,000 and where an investigation then needs to be conducted or a report filed. The system is inefficient, and organized criminals are learning how to evade detection through ever more sophisticated laundering techniques. Similarly, terrorist financing behaviour, often viewed as money laundering in reverse, is even harder to detect, as transactions often appear to be normal.

Regulators view transaction monitoring as a fundamental pillar to detecting and preventing illicit money flows, and advocate that financial institutions not only invest in sufficient staff to review alerts generated but continually monitor and tune the effectiveness of their monitoring tools. However, even with appropriate resources and tuning, illicit money can and will flow through the financial network. The question is whether there is a new paradigm in which false positives can be reduced while anomaly detection improved and if so, ...