AI: Key Anti-Money Laundering strategy

Artificial Intelligence (AI) is fast becoming European banks’ key defense against money laundering that finances terrorism and other criminal activities around the world.

Anti-Money Laundering (AML) is a key priority in the EU. The latest EU Anti-Money Laundering Directive (AMLD IV), published in 2015 and confirmed by the European Parliament, mandated higher safeguards for financial flows from high-risk countries and enhanced the powers of EU Financial Intelligence Units. Penalties are at an all-time high for banks transacting with known or suspected money launderers. The responsibility lies with banks to stop money laundering. The traditional approach to AML was largely manual and rules-based, focused on process not results. Guided by the principle, “Know your Customer” (KYC), banks followed step-by-step processes to check the identity of customers, their business partners and beneficial owners against the AML watch lists.

This rules-based approach involves tools that methodically flag cash transactions and transfers that are over a certain amount in Euros or involve certain countries. Human AML monitors then trawl through bank transactions manually and decide which should be further reviewed. This process is ineffective in identifying transactions designed to circumvent controls such as those involving complex layers of shell companies. In the last decade, a new generation of AI-driven AML algorithms has arisen. Enabled by hardware advances, distributed processing and Big Data, the new tools do not rely on strictly defined rules or parameters. Instead of looking at each transaction individually, they do holistic, contextual analysis to detect anomalous transactions, using broadly defined product, customer and risk type parameters. The algorithms “learn” dynamically--by recursively evaluating their own output, adjusting to changes in transaction activity and discovering new patterns for smarter future detection.

What has been the result of using AI-based technology? There has been a reduction in the volume of transactions flagged to be reviewed. Alerts are clustered by risk level, allowing analysts to better prioritize higher-risk alerts. Review of suspicious transactions is facilitated by technology-based tools, reducing the time required for manual examination, making bank compliance more efficient. In one test of this technology by a major bank, the number of false positive alerts was reduced by 35 percent. Categorizing flagged transactions by risk level improved the bank’s accuracy in identifying suspicious transactions by more than four times.

The use of AI and machine learning technologies promises to significantly increase banks’ proficiency in detecting patterns of money laundering, helping them stay at least a step ahead of terrorists and other financial criminals.