Financial institutions are reporting a dramatic increase in synthetic identity fraud — a scheme in which criminals combine real and fabricated personal information to create fictitious borrowers — and credit analysts say artificial intelligence tools are both enabling the attacks and serving as the primary line of defense against them.
The Northgate Financial Crimes Research Consortium released data Thursday showing that synthetic identity fraud accounted for an estimated 34 percent of all new account fraud losses at participating lenders over the past fiscal year, up from 21 percent three years ago. Total losses attributed to the scheme industry-wide exceeded several billion dollars, though precise figures vary depending on how institutions classify and write off disputed accounts.
“The profiles these operations construct are extraordinarily convincing,” said Carmen Whitfield, director of fraud analytics at the consortium. “They have credit histories, employment records, sometimes even social media footprints. A human reviewer following standard verification protocols would have no obvious reason to flag them.”
The growth of generative AI tools has made the construction of synthetic personas faster and cheaper, according to the report. Tasks that once required dedicated technical expertise — generating consistent personal histories, producing plausible supporting documents, establishing digital trails — can now be substantially automated, lowering the barrier to entry for criminal organizations and enabling them to operate at a scale previously accessible only to well-funded groups.
Fraud rings typically spend months cultivating synthetic identities before attempting to monetize them. They open small credit accounts, make regular payments, and gradually build credit scores high enough to qualify for larger loans. When the accounts are eventually charged off, the damage extends across the entire credit lifecycle — from origination through collections — making it difficult for institutions to identify the fraud point retrospectively.
Banks and credit unions have responded by deploying their own AI systems, trained on patterns of known synthetic fraud cases, to evaluate applications in real time. These systems assess factors that human reviewers cannot efficiently process at scale: the velocity of address changes across multiple applications, subtle inconsistencies in device behavior during digital sessions, and statistical anomalies in the clustering of credit inquiries around specific geographic nodes.
“It is fundamentally an arms race,” said Dr. Leonora Takahashi, a researcher specializing in adversarial machine learning at Crestfield University. “The offensive tools get better. The defensive tools get better. The question is always which side is learning faster at any given moment.”
Smaller community lenders say they are at a disadvantage because they cannot afford the most sophisticated detection platforms, which require large proprietary datasets to train effectively. Several trade associations have begun lobbying for shared fraud intelligence networks that would allow smaller institutions to benefit from the pattern recognition developed across larger portfolios without requiring them to build equivalent data assets independently.
Regulators have taken note. Banking oversight bodies in multiple jurisdictions have issued guidance encouraging institutions to update their identity verification standards, particularly for digital account opening channels, which fraud analysts identify as the most heavily targeted entry point. More prescriptive rulemaking is expected later in the year.
Consumer advocates have raised concerns about the collateral effects of more aggressive fraud screening. Individuals with thin credit files — recent immigrants, young adults establishing credit for the first time, people recovering from financial hardship — may face heightened scrutiny that results in higher rejection rates even when their applications are entirely legitimate.
“The system is chasing patterns, and some legitimate applicants happen to fit those patterns,” said advocacy attorney Simone Castillo. “There has to be a meaningful appeals process, and there currently is not one at many institutions.”
Financial institutions said they are actively working to reduce false positive rates, but noted that doing so requires data that institutions are often legally prohibited from sharing across organizational boundaries — a structural constraint that advocates and regulators are increasingly treating as a policy problem requiring a legislative solution.