NEW YORK—Artificial intelligence is rapidly transforming consumer lending fraud from an operational headache into what one industry research firm is calling a potential existential threat for financial institutions, as increasingly sophisticated fraud schemes overwhelm traditional controls and force lenders—including credit unions—to rethink how they evaluate borrowers.
A new report from Celent, commissioned by Zest AI, found 82% of U.S. lenders reported higher fraud losses in 2026, with more than one-third seeing double-digit increases. The report argues the surge is not a temporary spike but a structural shift fueled by generative AI tools capable of creating synthetic identities, falsified income documents and increasingly convincing loan applications that can bypass both automated systems and human reviewers.
For credit unions, which represented 43% of institutions surveyed, the findings underscore a growing concern that fraud losses are becoming intertwined with broader credit deterioration. Celent reported that 93% of lenders now believe fraud directly contributes to credit losses, as fraudulent borrowers increasingly appear creditworthy during underwriting before defaulting shortly after receiving funds.
“Fraud risk is not only the fastest-growing lending risk in 2026, it could lead to an existential crisis in the lending business if poorly managed,” the report states, warning that weak fraud controls can create losses that remain embedded in portfolios for years.
The study paints a picture of lenders struggling to keep pace with fraudsters that are now using AI to fabricate employment records, bank statements and even synthetic identities assembled from real and fake consumer data. Among the fastest-growing fraud categories cited were synthetic identity fraud, application stacking—where borrowers simultaneously apply for loans across multiple institutions—and “bust-out” fraud schemes in which borrowers initially appear legitimate before rapidly maxing out credit and disappearing.
Legacy Systems An Issue
Compounding the problem, many institutions are still relying on legacy fraud systems built for what Celent described as “yesterday’s fraud environment.” Fewer than one-third of lenders surveyed currently use advanced tools such as AI-driven fraud models, alternative data analytics or fraud consortium intelligence-sharing systems, despite widespread acknowledgment that existing tools are falling behind increasingly sophisticated attacks.
At the same time, lenders are throwing more people and money at the problem. The report found 70% of institutions are increasing fraud staffing in 2026, while 75% plan to increase fraud technology spending. But Celent warned that simply adding personnel is becoming unsustainable as digital fraud attacks scale faster than manual review teams can respond.
Instead, the firm argued the next phase of fraud prevention will depend on integrating fraud detection directly into underwriting and credit decisioning systems rather than treating fraud as a separate checkpoint after the fact. The report recommends combining fraud scores with credit analytics in real time, automating low-risk reviews, improving data quality and participating in cross-lender fraud consortiums that can identify coordinated fraud rings operating across multiple institutions.
The study pointed to a disconnect between industry awareness and actual collaboration. While 73% of lenders agreed fraud data-sharing consortiums benefit the industry, only 34% currently participate in one. Another 46% said they would participate if a suitable consortium existed.
For credit unions, the challenge may be particularly acute. The report noted that smaller institutions often lack the technology budgets and in-house expertise of large banks, though many share resources through CUSOs and cooperative partnerships. Still, Celent suggested the institutions that most successfully integrate AI-driven fraud analytics, automation and shared industry intelligence into loan origination workflows could gain a competitive advantage by reducing losses while speeding approvals for legitimate borrowers.
“The next phase of competitive advantage will come not from isolated tools but from integrated decisioning ecosystems,” the report concludes. “Institutions that embed fraud intelligence directly into loan origination will move faster, approve more legitimate borrowers, and protect portfolio performance more effectively.”