Ninety-three percent say fraud is now eating into their credit losses.
Not just costing them time and resources—actually contributing to loan defaults and charge-offs. And 82% report their fraud losses got worse in 2026 compared to the year before.
Those numbers come from a new Celent study commissioned by Zest AI, which surveyed 115 U.S. financial institutions. The takeaway? Fraud isn’t just an operational headache anymore. It’s a direct hit to profitability, and the traditional playbook isn’t working.
The Fraud That Hides in Plain Sight
So what’s changed? The fraud tactics themselves. Sixty-one percent of lenders called out synthetic identity fraud as their fastest-growing problem in 2026. Close behind: bust-out fraud at 56% and application stacking at 55%.
Here’s why these are particularly nasty. They’re designed to slip past the defenses of any single lender. Think about application stacking—someone applies for loans at five different institutions in a short window. Each lender runs their checks, sees nothing alarming in their own system, and approves the loan. Meanwhile, the borrower is racking up debt they never intend to repay, and no one institution sees the full picture until it’s too late.
Synthetic identity fraud works the same way. Fraudsters create fake identities that look real enough to clear standard verification checks. They build credit history, sometimes over years, then max out their credit lines and disappear. It’s not opportunistic—it’s organized and methodical.
“Fraud has evolved from a contained risk into a systemic threat that is cutting directly into lender profitability,” said Craig Focardi, Principal Analyst at Celent. “What makes this moment different is the nature of the fraud types that are driving losses. Synthetic identities, bust-out fraud, and application stacking are not opportunistic acts. They are organized, cross-institutional attacks, and no single lender has the full picture on their own.”
Fighting Alone in a Team Sport
The core problem? Lenders are fighting a networked enemy with siloed defenses. Fraudsters operate across institutions, but most lenders are only looking at their own portfolio data.
The industry knows this. Seventy-five percent of lenders are increasing their fraud tech budgets this year. Another 70% are adding headcount to fraud teams. But here’s the disconnect: fewer than one-third currently use AI and machine learning fraud models, alternative data signals, or consortium-based intelligence—the exact tools built to catch what traditional fraud controls miss.
And get this: 73% of lenders agree that fraud data-sharing consortiums would benefit the industry as a whole. Yet only 34% actually participate in one. That’s a massive gap between knowing what needs to happen and actually doing it.
The Data-Sharing Gap
Why the hesitation? The research points to a few things:
- Many lenders are waiting for the right option: Forty-six percent say they’d participate if the right consortium existed, including 25% who would join a cross-lender fraud signal consortium right now, and another 21% still weighing the benefits.
- Current tools aren’t cutting it: Sixty-four percent admit their fraud technology doesn’t keep pace with new fraud methods. Throwing more money at outdated systems won’t solve the problem.
“These findings reflect a broader industry reality: the cost of fighting fraud is rising, and many institutions are struggling to keep pace with increasingly sophisticated attacks,” said Mike de Vere, CEO of Zest AI. “Fraudsters are operating across institutions, and lenders are largely still fighting back within the walls of their own portfolio. The answer for lenders is shared intelligence that makes cross-institutional fraud visible before it becomes a loss, and that is exactly where we are focused.”
What Actually Works
The path forward isn’t subtle. Lenders need to see beyond their own four walls. That means embracing AI-powered fraud detection, leveraging alternative data signals, and—crucially—participating in shared intelligence networks that make cross-institutional patterns visible.
Zest AI’s Fraud Detection platform is designed for exactly this challenge. It identifies first-party and behavioral fraud at scale, catching more than 50% of malicious intent. It surfaces over 40% more first-party behavioral fraud with minimal manual review. And it keeps more than 80% of consumer loan applications auto-decisioned, so lenders don’t have to sacrifice speed for security.
Bottom line: fraud has gone networked. It’s time for fraud defense to do the same.
The complete Celent report, “Combatting the Rise in Lending Fraud with AI and Data Sharing,” is available here.
Related:
Commonwealth Credit Union and Zest AI Partner to Launch CU Lending Collective to Bring AI-Powered Lending to Small Credit Unions
Financial Institutions Race to Adopt Generative AI in Lending, with 83% Boosting Budgets in 2026