Tech company Oracle is developing anti-money laundering tools with gen AI amid climbing fraud and financial crime. 

Nearly $2 trillion — or roughly 5% of global GDP — is expected to be laundered this year, according to a Moody’s Ratings report published in June.

“Criminals are more incentivized … than investigators because, for criminals, it’s their livelihood,” Jason Somrak, head of financial crime products at Oracle Financial Services, told FinAi News. “For FIs, it’s like a cat-and-mouse game, always trying to figure out the next evolution of fraud.” 

To stay ahead, Oracle keeps an eye on criminals trying to scam the system by monitoring places where they share ideas, Somrak said. 

“There was a service on the dark web called ‘anti-analysis’ where criminals could submit a transaction, and it would tell you what’s the likelihood that a bank is going to detect you for money laundering or evasion,” Somrak said. “This kind of exercise allows us to come up with new controls and detect them better.” 

By studying evolving criminal activities and talking to customers, Oracle has created solutions to track and investigate money laundering attempts as well as and inform regulators, Somrak said. 

How it works 

Oracle’s AI model evaluates client patterns to flag suspicious transactions rather than just flagging large amounts of money changing hands, Somrak said, adding that if a vetted client has a history of moving large sums, the AI model will not flag those transactions. 

“With our behavioral AI model, FIs can reduce false positives by up to 50%, which leads to more efficient investigations and [suspicious activity reports] filings,” Somrak said. 

Oracle has deployed gen AI-driven models for investigating anti-money laundering transactions and onboarding clients in KYC processes, Somrak said. The model can also run background checks on new clients. 

“We have deployed gen AI as an investigation assistant,” Somrak said. “It can help investigators in finding essential information about a company or individual during the investigation process and synthesizing the information.” 

The AI investigative tool can perform nearly 80% of the knowledge-finding tasks for the human investigator, Somark said. But it’s the human who decides if the transaction needs to be filed with regulators. 

Oracle works with OpenAI for its gen AI efforts but is evaluating other models to make its offerings multimodal and more comprehensive, Somrak said. 

Check out our exclusive new bank industry data here.