Using Artificial Intelligence to Minimize Risk
Money laundering continues to be a serious concern for financial institutions across the globe, as legacy technologies allow an unacceptable number of illicit transactions to go unidentified. However, many forward-thinking financial institutions have begun working smarter by using artificial intelligence (AI), dramatically improving the effectiveness of their investigations, and enabling them to identify suspicious activity and high-risk customers and entities.
In response to increased regulatory scrutiny and related fines, covered financial institutions are spending billions of dollars annually to reduce the likely risk of enforcement actions being taken against them. The scope of regulatory action is evidenced by the U.S. Department of Treasury’s Office of Foreign Assets Control’s (OFAC) 18 civil monetary penalties in 2016 totaling approximately $27.8 million, and by the Treasury’s Office of the Comptroller of the Currency’s (OCC) 15 civil monetary penalties in in 2016 totaling almost $226.5 million. The volume of fines continues to increase year over year. In 2016, $42 billion in fines for AML noncompliance were paid to federal regulators by the largest North American and European financial institutions.
Regulatory actions also bring less obvious, yet no less costly remediation expenses in the form of legal fees, independent monitors and additional investigative staff. A court-appointed compliance monitor could cost a financial institution at least $3 million per month. Furthermore, the unwanted regulatory attention imposes reputational risk and potential damage to shareholder confidence. Despite increases in AML compliance spending by financial institutions, money laundering has continued unabated, and the risk of regulatory penalties has remained high. According to a WealthInsight report, global AML spending will exceed $8 billion in 2017, up from $5.9 billion in 2013.
The number of financial transactions will continue to increase globally, as documented by recent data from the Society for Worldwide Interbank Financial Telecommunication (SWIFT). The SWIFT data shows that an average of 27.45 million SWIFT messages for financial transactions were processed each day, a 6.3 percent growth over March 2016 according to SWIFT.
To combat money laundering, transaction monitoring systems (TMS) are relied upon as a first line of defense by financial institutions. However, TMS can allow an unacceptable number of illicit transactions, or false negatives, to go unflagged due to their dependency on the identification, development and manual implementation of new TMS rules and scenarios. Dogged innovation on the part of money launderers and practices such as below-the-line testing (BTL), which allows all transactions under a specific threshold to go unchecked, exacerbate the problem.
A rapid increase in the number of financial transactions and rules has led to an influx of flagged transactions. A 2013 report from Aite Group estimated that the number of AML alerts worked by financial institutions went from about 5.76 million in 2009 to an estimated 6.89 million in 2012. By 2016, the analyst firm estimated the number of investigations would have risen to about 10.36 million. This influx has led to high rates of false positive alerts produced by TMS. By some industry estimates, approximately 90 to 95 percent of the alerts generated by TMS are false positives. Since modern TMS operate on static, rules-based engines, they require frequent tuning, governance and oversight, which trigger more time-consuming and expensive investigations. However, merely increasing the number of AML investigators within banks and covered institutions will not resolve this problem.
Recent advancements in data science, including artificial intelligence (AI), machine learning and big data management, complement existing TMS to provide compliance teams with an end-to-end AML risk prevention, detection, investigation and mitigation platform. Financial institutions can benefit from these advanced solutions’ ability to analyze massive amounts of information from TMS, KYC databases, all lines of business (LOB) customer information and the deep web, as well as from investigative databases and internet sources.
Artificial intelligence has improved AML operations in several ways:
- Identifying illicit transactions that go unflagged by TMS
- Automating the investigation of all transactions including “below the line” transactions that currently go completely unmonitored
- Improving the ability to assess transactions or cases that historically have scant relevant data and lead to de-risking entire jurisdictions
- Continuously improve the AML process by identifying innovative and new money laundering techniques, as well as updating rules-based TMS
- Processing the flags created by TMS and scoring each in a manner that allows cases that do not warrant investigation to be discharged and escalating specific cases that require human analysis
- Rather than human investigators spending scarce time researching and digging for information on a case, AI automates the research process by serving up the salient data points required for faster and more accurate human determinations and regulatory support
Financial institutions have begun working smarter through the utilization of AI, dramatically improving the effectiveness of their money laundering investigations and empowering those financial institutions to identify suspicious activity and high-risk customers and entities. The only way to effectively avert money laundering risk is to incorporate these new technologies into the AML compliance framework. AI-powered solutions help financial institutions prevent criminals and terrorists from compromising the integrity of the global financial system while supplementing their existing investigative teams and driving down operational costs.