By Chris Harrison, CEO of Fideo Intelligence
The numbers coming out around synthetic identity fraud should concern everyone in financial services. Industry estimates now suggest US economic losses from synthetic identity fraud could reach $30- $35 billion annually, accelerating quickly as fraudsters weaponize AI to scale attacks.
While banks and fintechs have spent years strengthening controls around traditional identity theft and card fraud, criminals have been playing a different game entirely by manufacturing identities that blend fact and fiction.
A real Social Security number. A fake name. An AI-generated photo. The result is an identity that appears legitimate, passes KYC checks, builds a history over time, and earns trust. Then they disappear, often leaving significant losses behind.
Synthetic identity fraud doesn’t rely on stealing a complete identity. It relies on assembling one. By blending real and fabricated data, fraudsters create profiles that appear internally consistent across systems. Many synthetic identities are deliberately engineered to avoid early risk signals. They don’t trigger alerts. They don’t behave erratically. In fact, they often look like some of the most reliable customers an institution has.
That’s the tragic irony: Many synthetic identities outperform legitimate customers in the initial stages. They’re designed to be perfect right up until they’re not.
Generative and Agentic AI Changed the Scale and Speed
For years, creating convincing synthetic identities required time, coordination, and human effort across multiple steps. Generative AI has already disrupted that model by dramatically reducing the cost and complexity of fabrication.
Today, fraudsters can generate realistic and coherent biographical data and plausible documentation at scale. Generative models produce identities that are internally consistent and increasingly difficult to distinguish from legitimate consumers. What once demanded specialized expertise can now be accomplished with inexpensive, widely available tools.
Agentic AI pushes this evolution even further. Rather than simply generating artifacts, agentic systems can orchestrate entire fraud workflows. These systems can autonomously test onboarding flows, adapt behaviors in response to controls, manage identity lifecycles, and optimize success rates over time. Synthetic identities are no longer static creations. They can age accounts, vary transaction patterns, and respond dynamically to friction in ways that closely resemble real users.
The result is a shift from opportunistic fraud to industrialized operations. The barrier to entry has fallen, the quality and persistence of synthetic identities have risen, and detection has become materially more difficult. Controls designed to identify stolen or compromised identities struggle when the identity itself is manufactured, adaptive, and continuously refined by intelligent agents.
A Payments-Specific Risk Hiding in Plain Sight
After more than three decades of engineering systems across financial services, fintech, telecom, and law enforcement, one lesson remains: Fraud exploits gaps between systems.
An identity that passes onboarding also passes downstream checks. A clean transaction history may conceal synthetic buildup over months or years. Point solutions work well in isolation, but fragmentation creates blind spots.
This is especially true for payments, where systems prioritize speed and scalability. Decisions happen in milliseconds, with risk thresholds set to reduce friction. Synthetic identities exploit this setup by behaving precisely as legitimate ones would.
They tokenize cards. They create wallets. They transact predictably. They stay below velocity limits and avoid anomalies. To payment platforms, they look like trusted customers, sometimes even more so than real ones.
That’s what makes synthetic identity fraud particularly dangerous in payments. The risk isn’t just at onboarding. It’s embedded in authorization, provisioning, and ongoing transaction flows. By the time losses surface, the identity has often been trusted across multiple payment rails and partners.
Improving Signal Quality and Connectivity Across the Entire Lifecycle
As institutions confront this challenge, it’s critical to separate decision-making authority from signal quality and to recognize that both must operate across the full customer and transaction lifecycle, not just at onboarding.
Across the industry, there is growing recognition that outcomes in KYC, KYB, AML, sanctions screening, onboarding, provisioning, continuous due diligence and transaction monitoring all depend on the quality, consistency, and continuity of the identity and risk signals flowing through those processes over time. Synthetic identity fraud exploits breaks in continuity by assessing identity confidence once and then assuming it remains valid.
Improving identity intelligence does not mean relying on a single, front-loaded check. It means strengthening and connecting signals at multiple points. This can be during onboarding, as accounts age, as payment credentials are provisioned, and as transaction behavior evolves. When identity and fraud signals are continuously assessed and reassessed, institutions gain a clearer picture of whether trust is being earned or quietly manufactured.
As those signals become more precise and reliable across the lifecycle, organizations can improve match accuracy, reduce false positives and manual reviews, and deliver a better experience for legitimate users without weakening controls or adding unnecessary friction. The benefit is not just earlier detection, but ongoing confidence.
Crucially, this approach does not shift regulatory accountability or compliance responsibility at any stage. Final decisions remain with the financial institution, sponsor bank, or designated KYC/CIP provider. Strengthening identity and risk signals throughout the lifecycle ensures those decisions, whether made at onboarding, during monitoring, or at transaction time, are informed by better, more complete information.
The Questions Institutions Can’t Avoid
This moment demands hard questions—especially for banks, credit unions, fintechs, and payment platforms:
- How confident are we that the identities we trust today are real?
- What percentage of our portfolio or transaction volume could already be compromised by synthetic identities?
- Where do our controls stop—at onboarding, at authorization, or somewhere in between?
Juniper Research forecasts a “tidal wave” of synthetic identity fraud that will push global fraud costs to $58.3 billion by 2030. The question is: how will the fintech and financial services industry respond? Synthetic identity fraud isn’t loud. It doesn’t spike overnight. It blends in, performs well, and waits.
In a time when AI is manufacturing trust at scale, organizations must rethink how that trust is earned, validated, and maintained across the entire customer and payments lifecycle. Those who close the gaps will reduce risk without sacrificing experience. Those who don’t see the fraud coming until the damage is done.
About the Author
Chris Harrison is the CEO of Fideo Intelligence, and his leadership and direction have been instrumental in helping businesses safeguard trustworthy customer interactions. With over 30 years of experience designing and building intelligence systems, Chris has amassed invaluable knowledge of consumer behavior and identity. He combines his business management and engineering expertise with a proven ability to build high-performance leaders, teams, products, and architectures.
Chris began his career as an engineer and data architect, with formal training in database theory, systems design, programming, and artificial intelligence. Chris’s career has included roles in both the Government and the private sector. He has worked with financial institutions, including the largest banks in the world, as well as insurers, retailers, healthcare providers and a host of other entities across many verticals. His interest in protecting public trust has led him to create solutions that enable companies to engage consumers while ensuring the safety and security of interactions and data.
Previously, as the former President and CTO of the company, Epsilon, he led product, engineering and Data Science teams that went on to build award-winning products. He also led the professional services, sales, product marketing, and operations functions end-to-end.
Recent PaymentsNEXT news: