By Ray Birch
BOSTON—A number of financial institutions, particularly major banks, are turning to artificial intelligence (AI) to help them with their anti-money laundering (AML) programs.
The institutions are leveraging AI not only to improve how they sift through AML alerts to stay abreast of the latest threats, but are also using AI to make Suspicious Activity Report (SAR) submissions streamlined and faster, according to Aite Group. The banks, some of which are now in pilot tests, are also using AI to catch money laundering incidents and block them as they occur. Typically, the majority of money laundering incidents are discovered well after the transaction has been completed.
The moves come at a time when regulators are placing stricter rules on FIs, expecting them to not only catch but prevent more money laundering incidents. New customer due diligence rules are in place. In 2017, federal and state regulators stepped up AML enforcement, for example, and Citibank paid a $70-million penalty to the Office of the Comptroller of the Currency for violating its 2012 consent order relating to Bank Secrecy Act and AML deficiencies. Merrill Lynch paid $26 million, half to the Securities and Exchange Commission and half to the Financial Industry Regulatory Authority, for failing to detect and report suspicious banking activity involving billions of dollars in transactions.
Kristina Yee, senior analyst at Aite, says AI is needed to fight money laundering.
“The vendors that provide financial institutions with AML systems are pursuing varying levels of AI,” said Yee. “They are hiring AI experts and educating themselves. We are beginning to see pilots for AI AML systems at major banks.”
Streamlining Reports
Many of the AI tools being used and tested now are focused more on streamlining SARs reporting and helping teams sift their way through the massive number of the AML alerts that are sent their way, as well as the daily transactions.
“From a forms perspective, we are talking about using AI to prepopulate fields, what is still a very manual process, and also assisting with the investigative process,” said Yee. “That means looking at all of the alerts. The big banks are realizing that this is something they need to do, using AI to whittle down all the money laundering alerts that come their way to pay attention to those that are the most important to their business.”
Yee called the task of sifting through alerts a big issue facing financial institutions today. She said that one of the largest problems is falsely flagging a transaction that is legitimate.
“False positives are a real issue,” said Yee. “Many times it’s just grandma sending out Christmas cash and not someone trying to launder money.”
What is also time consuming, noted Yee, is a senior investigator who is skilled enough to look at an alert and in a few minutes determine it’s a false positive and move on, is the paperwork that follows that activity.
“It can take an hour or two to fill out the paperwork and explain why they did not flag this transaction,” said Yee. “The machine learning helps by pulling up the necessary data and prepopulating many areas of these forms. There is natural language processing and natural language generation programs that will even write part of the report.”
But what is slowing the adoption of AI for AML is not yet knowing what regulators will allow AI to do within the AML process. Will they accept SARs reports completed by AI? That is what is being sorted out today within many of the pilot programs, said Yee.
“Will all regulators be comfortable with a report that is not filled out by a human?” said Yee.
The next major step for AI in AML is to have machine learning prevent instances of money laundering as they happen. Yee said that involves training AI systems—learning customers’ and members’ behavior patterns, and creating safe lists of consumers as well.
Kristina Yee
Big Banks Leading The way
Yee said she believes AI used in this manner can reduce the number of false positives and stop instances of money laundering in real time. She said that only some of the very largest banks are testing this application of AI today.
“Yes, some of the largest banks and money transmitters have moved beyond using AI for automation of their AML processes to finding ways to use machine learning to be proactive, not reactive to fight money laundering,” Yee said. “This is much like the AI defenses now being used to prevent fraudulent credit card transactions. Financial institutions have traditionally looked at money laundering data well after the fact, a transaction could have been a week or two ago.”
While regulators have yet to formally convey their position on the use of AI for AML programs, they are expecting banks and credit unions to get better at catching money laundering incidents and preventing more of them form actually occurring, explained Yee.
Yee said that the decision to use AI for AML programs will not simply be based on if regulators approve the technology’s use, but also how many overall transactions the FI performs. She said that smaller institutions with skilled AML staff may still rely on their own talents to fight money laundering, since the number of suspicious transactions they may see are small. And, the price tag for the use of AI for AML is not small.
Yee said as more vendors introduce AI systems for AML that the price will come down. She also noted that any AI system that is cloud-based will also save the financial institution money, since they will not have the infrastructure costs. However, Yee noted that community banks and credit unions are facing a growing AML workload as marijuana is legalized in more states, and as more criminals turn to smaller FIs due to their traditionally weaker AML programs.