AI: The Double-Edged Sword in the Battle Against Financial Crime

Putting 'Scam Dens' Out of Business Means Using AI to Fight AI

Featurespace Chief Operating Officer Tim Vanderham told PYMNTS' Karen Webster in an interview that "when you think about the billions and billions of dollars that come from scams globally," the money made from illicit gains overshadows the revenues of some of the largest businesses around the globe. The conversation came against the backdrop of an article from The Wall Street Journal that detailed the rise of "scam dens," which operate essentially as business centers with sophisticated setups, complete with separate departments for training fraudsters, "onboarding" unwitting victims and KPIs used to determine whether certain scams are successful or not. Along the way, fraudsters are proving adept at using artificial intelligence to develop relationships and trust on the part of their victims, preying on human emotions and making off with individuals' life savings and retirement holdings, draining their bank accounts with brazen speed, notably through authorized push payments. In the United States alone, Vanderham said, the $2.7 billion in fraud reported just a few years ago represents only a fraction of the true tally -- mostly because people are embarrassed to report that they've become prey to unscrupulous scams. In the meantime, the crime syndicates are using the stolen funds to bankroll other crimes such as human trafficking and the drug trade. The banks and service providers tasked with battling fraudsters have a challenge when it comes to using AI to, well, fight AI. "They're not bound by the same criteria when it comes to leveraging AI and machine learning," Vanderham said. Financial institutions (FIs) are bound by ethical concerns and a burgeoning set of regulations that are still being hammered out. But the data that crosses the financial services system daily, and a collaborative approach to harnessing and analyzing that data can go a long way toward modeling what "genuine human behavior" looks like -- building profiles from individuals' trends and transactions, he said. Featurespace's models use behavioral analytics and collaboration to understand, for instance, how the transactional behavior of an individual consumer in London might differ from another individual living in South Africa -- or uncover whether a new transaction to Hong Kong might be a red flag if it comes from someone who's never transacted there before, Vanderham said. The data "helps banks and FIs with those warning signs," Vanderham said, which in turn fosters education and a reality check for end users so that they can go through extra validations to ensure that the transactions are warranted and are going where they should be headed. Featurespace has been investing in advanced algorithms to underpin fraud prevention efforts. Last year it launched TallierLTM, the world's first large transaction model, which uses generative AI to improve fraud value detection by up to 71%. "What OpenAI did around language and words, we've created for the payments environment -- modeling what genuine behaviors and transactions will look like," Vanderham said. It will be critical for the public and private sectors to work together to help regulations and technologies evolve. "We have to make sure that we're using advanced data algorithms and machine learning over this data to combat the fraud and to do everything we can to allow consumers to transact more freely," Vanderham noted. As he told Webster, "We're prepared to fight against these fraudsters -- to take them out, and to beat them at their own game" with AI and machine learning as two of the most prominent lines of defense (and offense) against such criminals.

Banks are using AI to stop the flow of trillions of dollars used in financial crimes through their networks

Financial institutions and regulatory technology firms are leveraging artificial intelligence to bottle the flow of cash being funneled towards illegal activities worldwide. An estimated $3.1 trillion in illicit funds passed through the global financial system last year, according to a Nasdaq's latest Global Financial Crime Report. Money laundering alone accounted for trillions of dollars that helped fund international criminal activities, including $346.7 billion in human trafficking, $782.9 billion in drug trafficking, and $11.5 billion in terrorist financing. Financial criminals are getting smarter and more dangerous with the help of advanced technologies that have become cheaper and easier to access than ever before. But financial institutions and RegTech companies are deploying many of the same technologies, including AI and generative AI, to help combat the growing criminal enterprise. "What AI is allowing us to do is to really start seeing how bad actors are interacting with others," said Nikhil Aggarwal, managing director in the Anti-Money Laundering Consulting practice at Deloitte Transactions and Business Analytics. "When you're able to visualize a broader network, you're able to do a deeper investigation into rings, and this really allows you to see some of those interconnected patterns in terms of how these threat actors are oftentimes working together." The U.S. Bank Secrecy Act was created in 1970 as a way to help financial institutions detect and prevent money laundering through their systems, also known as Anti-Money Laundering laws, or AML. Under the act, all financial institutions follow a set of guidelines known as KYC (Know Your Customer/Client) -- a process that these firms use to verify the identity of, and risks from, potential clients. Despite the regulation, financial crime has become more widespread with the rise of digital transactions, like online payments, withdrawals, and deposits. More than half of Americans use digital wallets more than their cards or cash, according to the results of a Forbes Advisor poll published last year. This has created huge amounts of information on transactions and customer behavior -- and that's where AI comes in. RegTech firms use the technology to leverage the massive stores of data collected by banks to fight financial crime more efficiently and precisely. "AI is good at analyzing larger scale of data and spotting patterns in a very large scale of data," said Dagan Osovlansky, chief product officer at Israeli software company ThetaRay. "If you speak with any bankers, they will tell you they have tons of data and they don't necessarily know how to use it in many cases." ThetaRay, which employs its own proprietary machine learning algorithms, takes a risk-based approach to targeting financial crime. Using a large swath of data points, the firm's AI learns the normal behavior of banking customers in what's known as "unsupervised learning," a type of machine learning that learns from data without human oversight. This allows the technology to spot anomalies based on behavioral patterns, rather than human instruction. The firm's financial crime detection platform is used by over 100 financial institutions, including Santander, Payoneer, and Travelex. It alone monitors more than $15 trillion worth of transactions using AI. Last week, the company acquired Screena, a cloud-based, AI-powered screening firm that compares potential clients with lists of sanctioned parties. The partnership is part of efforts to keep up with advancing technologies that Osovlansky believes are being adopted by criminals faster than financial institutions. "The real question, in my view, is, who's winning? And I'm not sure the answer is the answer that we would like to hear," Osovlansky said. "I think we are playing catch-up," he added. Although its still its early stages, the use of this technology has already resulted in a significant decline in false positives -- the flagging of normal banking activity as suspicious -- at a number of ThetaRay's partner banks, including in Santander's corporate investment banking division, Osovlansky said. Santander has used ThetaRay's anti-money laundering solution, which analyzes client data to detect anomalies that could indicate money laundering schemes, since the end of 2019. Other players in the RegTech space include Lucinity, an AI software startup based in Iceland that uses AI to provide firms with insight to improve their financial crime compliance. Some financial institutions, however, have their own in-house systems to use advanced technologies fight and improve their detection of financial crime. HSBC co-developed its AI system with Google to check for financial crime. The bank uses AI to monitor about 1.2 billion transactions for signs of financial crime across 40 million customer accounts each month, Jennifer Calvery, group head of financial crime risk and compliance at HSBC, wrote in a June blog post. It claimed to spot two to four times more financial crime than it did before, with 60% fewer false positives. JPMorgan Chase chief operating officer Daniel Pinto said at the bank's investor day in May that AI will make its KYC process, including customer onboarding and monitoring, up to 90% faster by the end of next year. That means processing 230,000 files with 20% less people. JPMorgan, the largest bank in the U.S., has been a leader in AI adoption within the banking world for years, with the highest volume of AI talent of all major global banks. The biggest challenge for financial institutions and their partners is data availability. Underlying data blocks, data quality, and data hygiene are a "perennial challenge" when it comes to stringing data together to deploy AI effectively, according to Deloitte's Aggarwal. The process of cleaning data, although time consuming and tedious, will give financial institutions and their RegTech partners more "powerful insights," he said. "I think this is where the opportunity is upstream to get some of the data fundamentals in place," he said.