AI is speeding up fraud

Fraud used to look like one bad login or one stolen card. Now a scammer can use generative artificial intelligence to write a convincing message, clone a voice, fake an identity document, and test thousands of variations in minutes instead of days. (bankinfosecurity.com) (sec.gov) That speed is already showing up in the numbers. The Federal Bureau of Investigation said Americans reported more than $20.9 billion in internet crime losses in 2025, and government-official impersonation complaints doubled that year as artificial-intelligence-assisted scams spread. (govtech.com) (nextgov.com) The trick is not just better fake content. Artificial intelligence also lets criminals run cheap experiments at scale, like sending 10,000 slightly different texts and keeping the one version that gets the most replies, the way online advertisers test headlines. (sec.gov) (bankinfosecurity.com) Financial companies cannot answer that with more pop-up warnings and more password resets. Naga Gauda, who leads cybersecurity, risk, and fraud architecture at Intuit, said the design problem is catching fraud quickly without adding so much friction that real customers abandon the transaction. (bankinfosecurity.com) That is why fraud teams are shifting from one-time identity checks to continuous signals. Instead of only asking “is this your password,” modern systems watch behavior like device history, typing patterns, and whether this payment looks like your usual Tuesday afternoon purchase. (bankinfosecurity.com) (nist.gov) The United States government is moving the same way. The National Institute of Standards and Technology updated its digital identity guidance in 2025 to address deepfakes, spoofed documents, and continuous risk monitoring, which is a shift away from treating identity as a single checkpoint at the front door. (identity.org) (enterprisesecuritytech.com) Banks are also using artificial intelligence on defense. Feedzai said more than half of surveyed fraud and financial-crime professionals reported fraud driven by artificial intelligence and hyper-realistic impersonation, while Mastercard said its bank clients are using real-time artificial intelligence tools to cut fraud losses and reduce false declines. (feedzai.com) (mastercard.com) False declines are the part customers notice. If a bank blocks your real grocery purchase because an algorithm got nervous, that is fraud prevention working like a smoke alarm that goes off every time you make toast, which is why security teams keep talking about precision and maintainability instead of just “more detection.” (mastercard.com) (bankinfosecurity.com) The next fight is over identity proof itself. Deepfake video, face morphing, and synthetic identities are getting good enough that a selfie check or a scanned document is no longer solid on its own, so firms are layering document checks, liveness tests, device signals, and behavior history together. (nist.gov) (infosecurity-magazine.com) (sec.gov) So the story is not that artificial intelligence created fraud. The story is that it compressed the time between “new scam idea” and “mass deployment,” which is forcing every company that moves money to build systems that learn faster than the criminals do. (bankinfosecurity.com) (plaid.com)