LegalWeek 2020: Leveraging AI to Combat Financial Crimes

Originally published in the Legal Executive Institute.

By Joseph Raczynski

NEW YORK — With increasing data flow among global financial institutions expanding, compounded by newly minted hybrid financial organizations in FinTech, managing these highly regulated transactions is progressively more complicated.

A panel at last week’s LegalWeek 2020 explored the anti-money laundering (AML) trends in 2020, discussed the state of legal artificial intelligence (AI) solutions in AML compliance and screening, and described a case study where AI has been successfully deployed in combating financial crimes.

Panelist Cassie Lentchner, Senior Counsel at Pillsbury Winthrop Shaw Pittman, kick-started the discussion by bringing up the topic of regulatory expectations. The premise she offered stated that those institutions which contemplate or currently use AI for compliance functions must employ a risk management fundamentals model. “Scoping is critical to assess the regulatory requirements and expectations,” she said.

Currently, AI can be used to automate repetitive tasks, aggregate information, cluster information into groups, and confirm or monitor information. Clearly, we are not at the point where this technology can surplant institutional judgement and decision making, just yet, Lentchner added.

Now, more regulators are wading in on how compliance departments use AI. In fact, we are at a point where some of these algorithmic applications require financial institutions to get prior written approval from regulators, said Tim Mueller, Partner in Global Investigations & Compliance at Guidehouse. During exams, Bank Secrecy Act (BSA) and AML programs are being examined, and financial institutions are expected to demonstrate the AI functions being use, explain how they work, and even defend the work product to the regulator. Some defenses have withered from the regulator’s findings and from AI bias.

Using Intelligent Segmentation

The panel further discussed an Intelligent Segmentation case study, diving into the pitfalls of current approaches to AML compliance and how AI solves for it. Frequently, normal banking behavior is identified as suspicious, and this includes large bank deposits and rapid money movement from one bank to another. Despite billions of dollars spent on traditional monitoring (which often ends with a human internal investigator), 95% of system-generated alerts are “false positives,” said Mueller.

The AI approach involves three phases: machine intelligence, subject matter expertise, and increased efficiency & effectiveness. In this new model, Intelligent Segmentation leverages AI and identifies segments using a collection of related datapoints beyond additional identifiers. This allows the machine to better identify groups that should be monitored together and those that should be split. Ultimately when parsing the data, the machine can see patterns or white space, thus eliminating false positives like rapid bank-to-bank transfers by trusted organizations which would have been triggered in pre-AI models and have been difficult to identify previously by human investigators.

Another AI-enabled AML compliance case study that Mueller cited is Behavioral Network Analysis. With most compliance departments, the typical approach is to monitor suspicious activity by deploying rule-based detection scenarios, for example, collecting if-then-else logical statements. These new AI-generated systems make it possible to create a “contextual monitoring” approach that leverages Behavior Network Analysis and allows it to review entities and their interconnected relationships and transactions. This has proven to be very effective, especially in trade and financial market businesses, explained Mueller.


Now, more regulators are wading in on how compliance departments use AI. In fact, we are at a point where some of these algorithmic applications require financial institutions to get prior written approval from regulators.


Another panelist, Dr. Sam Small, Chief Security Officer of ZeroFOX, said that based on his background in academic security research, he had uncovered the ease in which social media can be co-opted through impersonation, account takeover, piracy, threats of violence, fraud, and misinformation. Fortunately, he added, there are ways to counter it. Small’s AI approach dissects social media accounts by parsing the information within these accounts at large scale. In this way, the model can break down and characterize social media accounts into the following components: media content, avatars, name and username, text content, hashtags, dates, actions, and URLs.

Then, the model delves into each component: Are URLs safe? Are hashtags benign or seemingly suspicious? Each piece can be poured over, using highly tailored algorithms to uncover dubious activity and allowing the account users a way to fight back, Small explained.

The thrust of the panel’s conversation proved that AI will be omnipresent for compliance with AML and fraud amid the explosion of global financial data. The universal agreement among the panelists was that the guiding principles to need to gain acceptance from regulators, internal audit, and compliance department leadership include using proven, defendable technology, being transparent, and augmenting processes that you have in place.

The rapidly changing landscape of financial data compels compliance organizations to adopt these tenants or be willing to incur real risks for their company if they do not.

About Joseph Raczynski (141 Articles)
Joseph Raczynski Technologist/Futurist Joseph is an innovator and early adopter of all things computer related. His primary focus is around the future of technology, and speaks globally about Legal Technology, Cybersecurity, Blockchain, Artificial Intelligence, Cryptocurrency, and Robotics (drone technology). He also writes about the future of technology as it impacts the crossroads of our personal and professional lives, especially in the legal sector. Under the umbrella of Thomson Reuters, The Hearing a Legal Podcast, Joseph hosts a podcast with a technology focus. He also serves as a mentor with the Columbia | IBM Blockchain Accelerator and guest lecturer at Fordham University School of Law, as well as a Cybersecurity Committee Member at the University of South Florida. Joseph founded wapUcom, LLP, consulting with companies in web and wireless development. As a side project DC WiFi was created to help create a web of open wireless WiFi access points across cities and educate people about wireless security. Currently Joseph is with Thomson Reuters Legal managing a team of Technologists for both the Large Law, Corporate, and Government divisions in the US. Joseph serves the top law firms in the world consulting on legal trends and customizing Thomson Reuters legal technology solutions for enhanced workflows. He graduated from Providence College with a BA in Economics and Sociology and holds a Masters in eCommerce and MBA from the University of Maryland, Global Campus. You can connect with Joseph at JoeTechnologist.com or JosephRaczynski.com or onTwitter @joerazz

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: