Resistant AI: Unveiling financial crime with sophisticated technology

For example, let’s say you’ve applied for a life insurance policy. On the other end, a team of underwriters looks at your application information — age, habits, personal data, income, lab results, blood tests, a physician’s summary — and manually identifies your risk against a predetermined set of markers. That level of risk then dictates how high a premium you’ll have to pay. Some applications are rejected outright. Today, it costs most life insurance companies upwards of $500 just to reject a life insurance application. A technology like generative AI can come to the same conclusion regarding those risk factors based on the biomarkers in your application at a fraction of the cost. Of course, an approved application carries a higher financial risk for the provider than a rejected application. So while a company may be able to use technology that is only 70% accurate to reject applications, that rate won’t cut it for approving them. Regardless, even if just half the manual work is eliminated customers will be served more quickly and effectively at a lower cost to the provider. The immediate opportunity for industries like healthcare and insurance is to identify which applications they can accept under 100%, and begin to leverage generative AI for those tasks. Applied to the world of healthcare and insurance, generative AI has enormous potential capabilities. For customers of these industries, costs continue to go up and access to care is dwindling. In the United States especially, going to the doctor is an experience full of friction. Recently, I started to feel a bit unwell the day before a long flight. It wasn’t an emergency situation, but a time sensitive one. Though I’ve been seeing my primary care physician for over 20 years, they couldn’t see me on short notice and sent me off to urgent care. All I really needed was a quick diagnosis to give me peace of mind before I sat on a plane for 16 hours. A technology like ChatGPT has the potential to provide a diagnosis like that, without requiring a trip to the doctor. Or worse, urgent care. Access to healthcare is critical. I have better than average insurance, but still couldn’t get the help I needed when I needed it. Think of how it must be for one of the millions of people who don’t have the same quality of insurance or care. It shouldn’t be like that, nor does it need to be. Technologies like ChatGPT have the potential to democratize healthcare, to make it more available and accessible in a way that betters our collective quality of life. The big question for today’s incumbents is whether they’ll be the ones providing that improved experience, or whether they’ll go the way of Blockbuster. The opportunity to adapt and improve is at their fingertips, but action is imperative.

The ACAMS Annual AML & Anti-Financial Crime Conference in Hollywood, Florida, is a high-profile event in the anti-money laundering sector, serving as an annual platform for tech companies to demonstrate their solutions to FinTechs and traditional financial services.

At this year’s conference, Resistant AI, a previous double winner of the PwC Hackathon, showcased its application of artificial intelligence in combating financial crime.

The team at Resistant AI had over 2.3 million synthetic customer and transaction data points to analyse within a three-day period. Using their extensive experience gained from partnerships with leading payments firms and neobanks, they identified instances of fraud, money laundering, sanctions evasion, and other suspicious activities. What distinguished Resistant AI’s approach was the ability of their tool to put these instances into context.

Understanding the extensive network of money movements and account relationships is a challenging task, especially with millions of data points and a race against time. The team had to incorporate a wealth of additional information such as IP address data, contact numbers, physical addresses, and VPN information provided for the Hackathon. The ability to visualise and make sense of this complicated interconnected network, fondly named “The Tribble”, lies in the power of AI-powered forensic analysis.

Resistant AI’s approach is not a grand unified model focusing on specific AML tactics. Instead, it uses layers of simpler models that each highlight statistically unusual behaviours across different dimensions. This approach, which they call the ‘ensemble approach’, allows their tool to act as a highly accurate yet flexible countermeasure against uncommon activities.

The ensemble approach entails the application of various detectors that most effectively revealed hidden patterns and suspicious behaviours in the Hackathon dataset. These included anomaly clustering, detection of anomalous amounts, intelligent sanctions evasion screening, and more sophisticated analysis using large language models (LLMs) for adverse media screening, predicting transaction descriptions, and graph analysis of suspicious typologies.

This method enabled the AI to cut through the chaos of The Tribble and expose numerous fraud clusters displaying a range of suspicious behaviour. They included money mules, mirror trading, smurfing, anomalous high-value transactions, entity consolidation at a high-risk institution, suspicious transaction details, and sanctions/PEP evasion.

The team at Resistant AI showcased the power of AI at the conference by illuminating a case of suspicious activity involving a receptionist named Maggie Barelty. Their AI forensics detectors identified a series of anomalous high-risk actions that led to a detailed investigation of the account.

Read the story here.

Keep up with all the latest FinTech news here

Copyright © 2023 FinTech Global

Enjoyed the story? 

Subscribe to our weekly RegTech newsletter and get the latest industry news & research

Copyright © 2018 RegTech Analyst

Investors

The following investor(s) were tagged in this article.

RELATED ARTICLESMORE FROM AUTHOR