A customer with steady income, a clean history, and instant credit approval may not be the low-risk borrower they appear to be. As synthetic identity fraud becomes more sophisticated, banks need help spotting coordinated networks that manufacture the appearance of financial stability before those "perfect" applicants take out credit they never intend to repay.
The problem is evolving faster than defenses can adapt. Synthetic identity fraud, where criminals combine real and fabricated personal information to build credible-looking borrower profiles, is becoming more advanced as AI enhances the ability to create convincing manufactured documents, identities, and activity histories.
The stakes extend beyond lenders' balance sheets as undetected fraud contributes to tougher verification, higher costs, and greater friction for legitimate consumers seeking credit.
How synthetic identity fraud works and why it's hard to catch
Synthetic identity fraud differs fundamentally from traditional identity theft. Rather than impersonating a real person, criminals build an entirely new, fake persona by combining stolen details, including Social Security numbers and other personally identifiable information (PII), with fabricated information.
The sophistication lies in the coordination. When real information is combined with plausible fake information and spread across a coordinated network of synthetic personas, the traditional screening processes that banks use can completely miss the signs of network-based fraud.
This oversight is largely due to how current credit-risk systems assess applicants. Normally, banks evaluate applicants individually—a process that can fail to notice coordinated behavior that is occurring across a network of synthetic accounts. For example, one customer opening an account may look legitimate, and a second account moving money to the first may seem normal. But when dozens of coordinated accounts are sharing addresses, moving funds between institutions, or simulating income through circular money flows, the collective pattern reveals a fraud that individual screens rarely catch.
AI as both weapon and accelerant
AI is driving much of the acceleration and sophistication of widespread synthetic identity fraud.
Attackers can now use generative AI to quickly create fake identities and documents and produce convincing fabricated activity histories. They can even simulate human behavior on verification screenings and generate deepfakes to pass biometric tests.
These systems also enable rapid refinement of these scams with fraudsters testing attack patterns and iterating continuously to improve, and expand on, successful tactics.
All of this comes as Sumsub reported early last year a 311% increase in North American synthetic identity document fraud between 2024 and 2025, and Deloitte projects synthetic identity fraud will generate at least $23 billion in losses in the U.S. by 2030.
A real-world case study: When signals hide in plain sight
When working with Teradata, a major Asian bank uncovered exactly this type of network-wide synthetic identity fraud in action.
Synthetic accounts that had been previously opened were applying for new credit, and on the surface, everything appeared normal. In reality, criminals were manufacturing the illusion of solid income by moving money across accounts, creating the appearance of reliable earnings. As with many banks, their credit risk models didn't flag the behavior as suspicious.
As a result, entirely manufactured "customers" were being approved for new credit lines. The turning point came when the bank shifted its analytical lens. Using graph-based analytics with Teradata to map these relationships, the bank identified tightly connected clusters of accounts and circular money flows that were simulating income.
The shift allowed the bank to see patterns that were previously invisible at the individual level but became unmistakable at the network level. Once the bank could see these patterns, it saved millions of dollars in potential losses and strengthened its real-time lending decisions.
The data access barrier
Another layer to financial institutions' ability to fight this type of fraud lies in data access issues. A 2026 study by Teradata showed that 50% of financial-services leaders report that governance, security, or access restrictions fundamentally limit AI agents' access to enterprise data.
AI-enabled fraud detection depends on connecting behavioral patterns across accounts, geographies, and transaction types. But when data is siloed, or spread across legacy systems and different departments, the fraud patterns that graph-based analytics can reveal become hidden.
To fight these types of synthetic identity scams, banks must first connect these historically siloed data systems in order to effectively understand these signals at the network level and act on them in real-time.
The consumer’s role in prevention
For consumers, it's imperative to stay vigilant about personal data that may be used in AI and network-based scams.
Consider these essential protections:
- Freeze your credit files to prevent unauthorized account openings
- Regularly review reports from credit reference agencies for unauthorized activity
- Set account and credit alerts to catch suspicious changes quickly
- Be careful sharing sensitive personal information, especially national ID numbers, such as SSNs and identity documents
- Monitor for signs of fraud, including credit inquiries or accounts you don't recognize
Solving a growing threat
Synthetic identity fraud is no longer a niche threat—it's an industry-wide challenge with real financial consequences. As AI makes it easier and faster for criminals to mount and refine attacks, banks must take equally sophisticated, AI-based action to spot these coordinated efforts before they can cause significant damage. For consumers and institutions alike, it's time to act now.