The mortgage banking industry has been slow to adopt AI. But criminals love it and are already making it harder to detect fraud.
Frank Abagnale believes that artificial intelligence is making it faster and simpler to commit financial fraud than ever before.
Abagnale, whose exploits were featured in the Steven Spielberg film "Catch Me If You Can," has since spent 50 years helping the FBI spot fraud. He freezes his credit, doesn't write a lot of checks, and doesn't keep a lot of apps on his phone. But protecting yourself from modern scams must go far beyond that, he said.
At a Mortgage Bankers Association conference in Washington, DC, Abagnale told business leaders that criminals are already making it harder to detect financial fraud.
"Verification must replace trust. In the environment today, you absolutely, positively need to know who's on the other end of that device before you send them any money or information," Abagnale said.
"Fraud today is less about breaking systems and more about influencing people."
Abagnale said that considering the long history of data breaches among business leaders and government agencies over the past few years, it's easiest to assume that everyone's identity has already been stolen. And those who steal the information package it and resell it years later.
"What used to take hours and printing presses to create ... today is done simply sitting at a computer, with AI's help," Abagnale said. "The difference today is I can send 10 million emails, and I'm still looking for 0.1% return."
Synthetic identities
Robb Hagberg, executive director of home lending fraud strategy for Chase Home Lending, said customers who are less experienced with technology are most at risk of the new fraud.
"If you think for one minute that you know more than fraudsters do about your product controls, you are sorely mistaken," Hagberg said. "They know what you're doing better than you know what you're doing. It's their job to manipulate, to work around the controls you put in place."
The technologies allow criminals to create more sophisticated imitations of people, like voice impersonations and deepfake videos. They can also deploy phishing emails and other scams at a faster scale than they ever have.
In 2024, fraudsters persuaded a Hong Kong worker to wire $25 million with a deepfake video purporting to be his boss. Wire fraud scams are always a major concern in real estate purchases, as scammers use social engineering and other tools to try to steal deposits and transfers during the transaction process.
The American Land Title Association said this month that 59% of firms reported seller impersonation fraud attempts, more than double the year before.
Amanda Padd, chief revenue officer of Secure Insight, said AI allows bad actors to defeat traditional security measures on a wide scale.
"Identity is just becoming that much harder to trust," Padd said. "A callback is still a useful control, but I don't think it's enough anymore. Don't eliminate human controls."
That’s why buyers, sellers, and real estate professionals need layered security. Effective protection combines digital safeguards—like passkeys, biometrics, and institutional controls—with multisource verifications and real-world interactions like phone calls and face-to-face meetings.
"What keeps me up at night is how scalable and sophisticated fraud can get to at some point," Padd said. "Not being able to stay ahead even though we have proper controls and processes and technology. That's pretty scary."
AI on the beat
MBA found, though, that the mortgage banking industry itself has been slower to adopt AI, and a study released by the organization on Monday found mortgage lenders and servicers are hesitant.
Their main concern is regulations and uncertainly about how those laws might change. Fannie Mae, for instance, determined lenders own their vendors' AI outcomes. Colorado just passed an AI act that holds businesses responsible for the outcomes of their AI use, too.
Executives think it's going to work alongside the existing tools. But the relationship part of the business remains that much more important.
Mandy Phillips, head of compliance for ACES Quality Management, was skeptical that AI would displace some people-focused workers, like loan officers. But it can also make other parts of the business more productive.
"I do picture it more as a tool to take our compliance people and our QC analysts and superpower them," she said. "But it will not replace us; it's a tool."
Abagnale said AI can also counteract the new frauds, and it has proven capable of analyzing fraud patterns in large datasets and detecting it in minutes. That's important because only a small fraction of fraud is detected today, due in part to how long it can take to detect.
Hagberg added that giving employees more experience in AI use helps them better identify fraud, as well.
Still, AI introduces its own problems, especially with risks of data leakage and making mistakes at scale.
"It can hallucinate with extreme confidence," Marty Allred, chief compliance officer of American Pacific Mortgage Corp., said.