The Green Sheet Online Edition
May 11, 2026 • 26:05:01
Combatting fraud in the age of AI
Cyber-enabled crimes defrauded Americans of nearly $21 billion in 2025, with cryptocurrency and artificial intelligence-related complaints among the costliest, according to the latest data from the FBI. The bureau reported that AI accounted for 22,364 complaints it received last year, costing Americans nearly $893 million.
Separately, Certik, a large blockchain security auditor, reported that crypto ATM scams in the United States jumped 33 percent to total $333 million in 2025 as AI deepfakes supercharged fraud.
Meanwhile, the LexisNexis Risk Solutions Cybercrime Report revealed that bot attacks were up 59 percent in 2025, with growth targeting ecommerce, gaming and gambling. Agentic commerce was still relatively new, so only small volumes of such traffic is reflected in the data. "But traditional automated (bot) traffic was evolving, with evidence of more sophisticated simulation of human behavior," the report stated.
"Breaches are at an all-time high," said Ken Michie, chief technology officer at Fideo Intelligence, a Denver company that specializes in fraud prevention and detection. "Generative AI makes it super easy to build new exploits. And data is going to continue to be more exposed." Michie added that significant "novel data"—data points and data sets—is showing up on the dark web.
And fraudsters are using that data to create synthetic identities, which enable them to orchestrate account takeovers faster than ever before. While the creation of synthetic identities and account takeovers is not new, what is new is the ease and sophistication by which it happens, Michie explained. Fideo, Michie noted, works with companies that monitor the dark web and can identify where bad actors hang out. "It's important to have good partnerships on that front," he said.
In February 2026, Fideo introduced an API developed specifically to ferret out fraudulent activity in real-time payments environments. Called Verify for Payments it includes such features as the ability to deploy across onboarding, merchant vetting, account linking and transactions; sub-second response times; and higher approval rates for legitimate customers.
"Payments fraud is evolving faster than traditional controls can adapt," Christopher Harrison, Fideo CEO, said in a press release. "Synthetic identities and AI-driven attacks are designed to exploit gaps between onboarding, compliance and fraud systems. Verify for Payments introduces a real-time identity intelligence layer that strengthens those systems upstream, enabling organizations to detect fraud earlier, approve more legitimate customers, and reduce friction across the payments lifecycle."
Clones that look and sound real
AI tools used by scammers are extremely sophisticated. Scammers are able to clone voices and generate believable IDs and realistic-looking deepfake videos, making impersonation harder to detect. These scams combine familiar voices or faces with personalized details and a sense of false urgency, for example, to scare victims into sending money through bitcoin ATMs ostensibly to avoid arrest or to bail a loved one out of jail.
It's not just individuals being scammed. An overview of the 2026 Association for Financial Professionals Fraud and Control Survey stated that "although deepfake technology is not as pervasive [as other types of financial fraud perpetrated on corporations], its emergence poses significant risks; manipulated audio and video can convincingly mimic authentic communications and undermine organizational security."
The 2025 State of Fraud and Financial Crime in the United States, a report produced by PYMNTS Intelligence in collaboration with Block, portrayed an industry being shaped by AI and behavioral analytics designed to meet more complex risk environments. And a more treacherous one. "Confronted with smarter fraud schemes, strict regulation, and a payments landscape evolving in real time, financial institutions are reaching an inflection point," the report stated.
But it is worth noting that while the payments landscape is evolving rapidly, creating new opportunities for fraud, many fraudsters continue to rely on tried and true ways of defrauding individuals and companies.
"What we've seen, unfortunately, is a rise in check fraud," said Ati Azemoun, vice president for business development at ParaScript, a software company that uses AI and deep learning to detect check and signature fraud. "The fraudsters have figured out that they can use AI tools to create fraudulent checks better. And they can do it at scale."
In fact, the AFP report revealed that checks continue to be the payment method most frequently impacted by fraud. Fifty-eight percent of organizations surveyed by AFP reported being targeted by check fraud in 2025. This compares with 30 percent reporting incidents of ACH debit fraud and 25 percent experiencing wire transfer fraud. Commercial card fraud was experienced by 21 percent of firms while fraud involving faster payment rails was reported by just 2 percent. Data published recently by the Federal Reserve suggests financial institutions are experiencing similar rates of fraud. One set of data points unique to the Fed report involves debit card fraud. The Fed described attempted fraud involving debit cards in card-not-present environments as "nearly universal."
Turning the tables on fraudsters
AI doesn't just benefit fraudsters; it can also supercharge fraud detection. "What AI does is it allows for adaptive learning," said Jason Schwabline, chief commercial officer at CheckAlt, a technology company that provides integrated receivables solutions. "Every fraud platform has gotten better because we've had to teach it."
Historically, fraud detection has been a batch process that is largely reactive. As the financial services industry moves to real-time adaptive learning and faster payments, however, fraud detection becomes proactive, which improves the accuracy and speed of identifying fraud. "We begin to guess scenarios as they come into play," Schwabline said.
'Not a bolt on'
"AI is not a bolt on," said Panhaj Sarda, senior vice president for product management at financial technology firm CSI. "AI has become embedded to help prevent and detect fraud." That's not to suggest it's static, however. "You have to test it on an ongoing basis."
Sarda uses the example of check fraud detection to illustrate his point. Mobile deposit has compressed the check clearing process, rendering the check system more susceptible to fraud. "We're working with banks implementing technology to detect check fraud as it is happening in real time," Sarda said. The technology examines the biometrics of the check itself. "This is the power of AI."
ParaScript's Azemoun described what he sees as a "generational shift" in the application of fraud detection. But there also has been a shift in how FIs identify fraud, and AI plays an important role in that.
For example, a common check fraud scam involves taking a legitimate check (more often than not one that has been intercepted from the U.S. Postal Service), washing (treating it with chemicals to erase the payee and amount information), and generating multiple copies with different payees and amounts. Then they funnel the altered checks through mules into their desired accounts.
To help Fis better spot these bogus checks ParaScript developed alteration detection software that can be deployed at the point of presentment. Tools also have been developed to identify deep fakes at account openings or other situations requiring signatures. "The gesture of signing is unique. You can classify it as a form of biometric," Azemoun said. "The pressure at which you apply a signature; the speed and fluenCy of how you write. Nobody else can really replicate that." Some companies even offer video tools that can analyze a person’s gait to determine whether they are who they say they are, Azemoun said.
Card brands deploy smarter risk management
Adaptive learning and the ability to spot anomalies in real time make AI a must-have tool for the card brands and their partners. "Our advantage lies in deploying AI not just to catch fraud as it happens, but to predict and prevent it," said Paul Fabara, chief risk officer and client services officer at Visa.
Visa revealed in a December 2025 press release that its AI-powered security systems identified and blocked 144 percent more suspected fraudulent activity over the Black Friday through Cyber Monday holiday weekend globally than it did over the same period in 2024. The company said it had invested $13 billion over the previous five years to build a "multilayered defense infrastructure" capable of analyzing transactions at "lightning speed."
Mastercard noted in a September 2025 report produced in partnership with Financial Times Longitude that 42 percent of issuers and 26 percent of acquirers had stopped more than $5 million in attempted fraud thanks to AI.
Last year, Mastercard began using a threat intelligence tool that leverages its acquisition of Recorded Future. The tool "provides customers with actionable, real-time and targeted risk insights to disrupt fraudulent transactions, inform strategic defense and ultimately enable a more proactive approach," said Johan Gerber, executive vice president of security solutions at Mastercard.
Mastercard said data from the tool had already helped identify and take down malicious domains that had impacted nearly 9,500 ecommerce sites and that were linked to an estimated $120 million in fraud.
"Effective cybersecurity will increasingly depend on threat intelligence that crosses sectors and regions. Vendors with broad transactional data and analytics will play a key role in improving information-sharing among financial services, merchants and across the industry," said Tracy (Kitten) Goldberg, director of cybersecurity at Javelin Strategy & Research.
Patti Murphy is senior editor at The Green Sheet, president of ProScribes Ink (www.proscribes.net) and self-described payments maven of the fourth estate. Her Today in Payments reports are a regular feature of the Merchant Sales Podcast.
Notice to readers: These are archived articles. Contact information, links and other details may be out of date. We regret any inconvenience.
