The Many Faces of Artificial Intelligence in Bank Fraud

AI has dramatically lowered the barrier to entry for fraudsters, who are now attempting attacks like account takeovers (ATO) and synthetic identity fraud at unprecedented speed and scale, resulting in billions of dollars in losses for financial institutions. At the same time, banks and credit unions are using AI tools for anti-fraud defenses like real-time transaction monitoring, behavioral analysis, and anomaly detection. 

It’s an intense, ongoing battle between AI-savvy criminals and banks, who view the technology as a key component of their anti-fraud strategies. The stakes are high: a Nasdaq Verafin 2026 global financial crime report found that fraud scams and bank fraud schemes totaled $579.4 billion in losses globally in 2025, with an annualized growth rate of 9.2% since 2023. 

AI has certainly played a big role in fraud growth. Matt Overin, Manager, Risk Management, at Logix Federal Credit Union, says that in the past banks’ fraud teams could spot scams easily and almost immediately because of simple mistakes like misspellings. But AI now offers a roadmap to producing polished emails and even how to move money after a successful scam. “Now you can just ask AI how to do the scam, and it will give you  
step-by-step instructions,” he says.  

The Nasdaq Verafin report highlighted the booming use of AI by bad actors, noting that criminals have quickly adopted “the latest advances in AI to perfect their scam playbooks.” What’s more, 90% of respondents cited an increase in AI-driven attacks over the past two years, while a different report found that AI fraud surged by 1,210% in 2025 alone.  

It should come as no surprise, then, that chief risk officers participating in ProSight’s 2026 CRO Outlook Survey cited AI-exacerbated fraud and financial crime as their second largest risk, with respondents noting that bad actors now have more potential entry points into banks and customer accounts because of an increased focus on digitalization and AI.  

AI, however, is also perceived as the greatest opportunity to get ahead in the fight against fraud. Three-quarters of respondents in the Nasdaq Verafin report said they expect to beef up their use of AI for financial crime protection, while the world’s largest banks plan to increase their spending on AI technologies by 20% this year. 

As scams become more sophisticated and prevalent, banks need to consider not only how they can employ AI tools but also what other strategies—like training and information sharing—they can use in the fight against AI-driven fraud.  

AI-Driven Risks 

Fraudsters no longer need to be master hackers. Instead, they can use AI-supported attacks to access customers’ personal data (passwords, credit cards, etc.), via, for example, phone calls that clone the voices of bank executives or highly personalized phishing messages that sound authentic. What’s more, thanks to AI, they can execute attacks rapidly (think hours instead of weeks) and on a massive scale that significantly increases the fraud management challenge for banks. 

One major threat that is growing significantly is account takeover fraud, where criminals use AI-enabled tools to manipulate customers and gain unauthorized control of their accounts.  

Overin says that ATO is the most prevalent type of fraud that Logix faces. “We’re seeing account takeovers where they’re hitting us and our members hundreds of times a day with fake phone calls and/or text messages,” he says.  

One important component of this type of fraud is AI voice cloning, which begins when scammers call up executives at financial institutions to try to record snippets of their voice. Overin has experienced this first-hand in the past, when he’s answered calls (seemingly from colleagues or members) and heard silence on the other end. After secretly recording Overin’s voice and using AI to clone it, fraudsters then call Logix members and pretend to be him. “Members will actually think they’re speaking to me,” he says. 

This is part of the “spray-and-pray” approach fraudsters use to execute rapid attacks. Complicating matters further, scammers often try to execute attacks after hours, knowing that Logix doesn’t offer 24-7 customer service. Feeling intense pressure, with no opportunity to speak with a customer service representative to verify false information, some members will give out their username, password, and credit card numbers in their eagerness to stop perceived fraud. But if a successful takeover is executed after hours, a member can get locked out of his or her account and will not be able to rectify the problem until the following morning, when Logix opens for business.  

Logix has also seen a large increase recently in synthetic identity fraud attempts, particularly through its online channel for new memberships. Overin says that it’s difficult to combat this type of fraud, because, after using AI to create a fake ID and build a credit profile, a scammer “just disappears” after, say, taking out a large loan and/or maxing out credit cards.  

Practical Anti-Fraud Uses of AI 

Financial institutions are learning how to use AI for anti-fraud purposes like detecting suspicious payments and behavioral anomalies, monitoring transactions in real time, and uncovering stolen identities.  

As part of its effort to prevent fraud, Logix has been using an AI tool from a document authenticity vendor for the past two years. The tool reviews the authenticity of documents for Logix member applications and loans.  

Bank statements, proof of residency, pay stubs, and W-2 forms are among the type of documents that the system can review to determine whether any alterations or additions have been made. “We’ll drop a document in,” Overin says, “and it will let us know whether the information is legitimate.”  

Through these reviews, Logix can determine whether an employer identified by a member candidate is a real firm or just a shell business. In addition, the tool can figure out if the candidate is actually employed or is simply using, say, an AI-created pay stub.  

Overin says the tool has already provided a huge return on investment. “We’ve saved literally millions of dollars each year in loan fraud alone,” he says. 

Logix has also seen a large ROI from the anti-spoofing technology it has recently implemented. That system has access to all the credit union’s phone numbers and is connected to a Logix server. Whenever a call to a member seems to originate from a Logix extension, the server communicates with the anti-spoofing system for a rapid verification check. The call then is either immediately terminated (if the system determines that it did not come from a Logix extension) or connects.  

The credit union’s future plans include more anti-fraud AI deployments. For example, it expects to integrate the document authenticity tool with its in-house consumer loan system, making it easier for the firm to track new account openings.  

Moreover, Logix is currently seeking a vendor tool that will allow the institution to use AI to write and test rules for its fraud investigations team. The vendor it chooses is expected to evaluate the firm’s existing anti-fraud measures and provide guidance on whether Logix needs to enhance its fraud mitigation—either through adding new systems or improving its existing anti-fraud tools.  

The Importance of Training and Information Sharing 

Offering training is another way banks and credit unions can keep up with scammers and protect their customers from AI-driven fraud. Today, this takes many different forms, including educational (in-person) forums, commercials, webinars and videos. It’s an effective way to teach about the latest scams, including AI voice cloning and synthetic identity fraud. 

Logix employs an AI-enabled “efraud prevention” platform to deliver training to its members. When, say, a member falls for an AI scam, the firm will describe the fraud on this platform and try to use it as a teaching lesson. Employees, meanwhile, have access to both an internal fraud training library and a public-facing learning module that uses AI to provide hundreds of online self-help videos—including training in AI fraud. 

Members of the credit union have access to that module, and can also benefit from Logix’s so-called Fraudometer, an online tool that asks customers about their day-to-day activities, provides a fraud score, and describes steps they can take to mitigate fraud.  

Information sharing is yet another way to combat AI-driven fraud. The ProSight Fraud Alert Network, for example, is a collaborative, authenticated, anti-fraud platform that provides members with insights, intelligence, and best practices.  

Overin says that Logix belongs to two different anti-fraud working groups—one consisting of 50 credit unions and another with 500. They hold virtual meetings where executives talk about the different types of AI fraud they are experiencing and share ideas for mitigation.  

Scanning the Horizon  

The fraud landscape has changed dramatically, with a notable shift from attacks on banks to attacks on customers. Previously, so-called social engineers would call a bank center to try to break into an account. Now, in their AI-fueled efforts to steal information and take over accounts, fraudsters are frequently attempting to scam customers of banks or credit unions directly.  

This year, ATO fraud figures to become even more prevalent with the emergence of agentic AI—autonomous, proactive AI tools capable of independently setting goals, creating step-by-step plans, and executing actions to solve complex problems.  

Overin says that it’s likely the industry will see an increase in ATO attempts because of agentic AI’s ability to execute scams rapidly and at scale. On the plus side, though, Logix’s fraud investigations tool will use agentic AI to generate low-level fraud alerts. “That will definitely help my staff focus on the high-level, high-risk alerts,” Overin explains. 

In 2026, he says, Logix will also continue to keep a close eye not only on the “emerging threat” of synthetic identity fraud but also on data breaches—because any time there is a breach, the credit union sees a surge in online  
fraud attempts.  

One additional emerging trend worth monitoring is fraudsters’ increasing targeting of commercial credit customers with AI scams. Commercial credit is not subject to as many risk controls as retail banking and may not have a dedicated fraud team, making it a more inviting target for scammers.  

AI fraud will undoubtedly continue to present a multitude of threats. To keep their customers safe, mitigate risk, and protect against huge losses, banks and credit unions will use a combination of anti-fraud AI tools, training and information sharing.