The Weakest Link In Fraud Prevention

In Catch Me If You Can, Leonardo DiCaprio brought to life the exploits of notorious conman Frank Abagnale. He forged checks, fabricated identities, and talked his way past bank employees, swindling millions before he was old enough to rent a car.

Sixty years later, one would expect that this type of forgery would be easy to detect – but the truth is far from it.

Artificial intelligence has turbocharged the problem. AI-assisted manipulation has made document fraud faster and harder to detect, enabling synthetic identities and convincingly altered IDs, bank statements, and pay stubs to be produced at volume. Digital document forgeries now make up 57% of all document fraud, surpassing physical counterfeits for the first time. And human ability to identify AI-generated content is no better than a coin toss.

As a result, documents have become one of the most exploited entry points for financial crime.

Meanwhile, regulators are demanding tighter controls, clear audit trails, and defensible decisions.

So how do lenders fight back?

We sat down with Jon Knisley, AI Strategist with global intelligent automation company ABBYY, to ask him a few questions about this urgent and growing problem. And more importantly, what actions lenders can take to defeat it.

QUESTION: Firstly, Jon, tell us a bit about document fraud and why it’s getting so much attention right now?

JON KNISLEY: As you’re aware, document fraud is the act of deliberately altering, using, or creating fake, or genuine, documents for deceptive purposes such as obtaining a loan or property.

The reason it has become a more urgent issue is due to the sophistication of fraud, with artificial intelligence being at the forefront of manipulation, in particular generative AI. This is leading to far greater monetary losses for financial organizations. In 2023, fraud scams and bank fraud drove an estimated $485.6 billion in global losses, while financial institutions spent up to $30 million per year on KYC alone. At the same time, financial crime compliance costs across North America and EMEA have climbed to $146 billion with regulators showing little tolerance for failures.

Lenders are challenged to balance low-friction customer service with stronger controls, even as onboarding becomes faster and increasingly mobile-first. What’s changed is not just volume, but capability. AI has dramatically lowered the barrier to document fraud, enabling fake identities, cloned templates, and skillfully edited bank statements and IDs to be produced at scale. One industry analysis found synthetic identity document fraud increased more than 300% in North America year-over-year.

The banking and financial services industries are the most impacted, followed by insurance, healthcare, and government departments. Yet most organizations still rely on fragmented tools and rules-based checks designed for a simpler era. This has pushed document forensics from a niche capability into foundational control. 

QUESTION: What are the different types of document fraud?

JON KNISLEY: Documents are now being presented in a variety of ways and of varying quality. They come in a range of file formats, such as PDFs, scans, photos, and word docs. Criminals rely on this variety when producing these fraudulent documents. Then institutions must sort through these unclassified documents at scale, fueling the ever-increasing challenge of fraud detection.

Document fraud is generally carried out with two types of documents: entirely false documents and manipulated genuine documents.

The false documents are completely fabricated with no basis in legitimate records. Fraudsters create these documents from the ground up, attempting to pass invented information off as real. Criminals may create or use false documents through techniques like:

1. Document forgery: When a document is designed from scratch to mimic an authentic one, often replicating layouts, formatting, and seals used by trusted institutions.
2. Template fraud: One of the oldest examples of cybercrime is the exploitation of editable templates. Criminals use these editable templates to create convincing replicas of official documents.
3. Synthetic Identity Fraud: This type of fraud typically involves combining real data elements (like a valid Social Security number) with fake information to create an entirely new, but fictitious, identity. This new identity can then be used to open accounts, obtain services, etc., creating a whole chain of fraudulent events tied to the false identity.
4. AI-Generated document fraud: With the rise of consumer-grade generative AI tools, fraudsters can now prompt software to create sophisticated-looking documents from scratch. This lowers the barrier to entry for inexperienced criminals, now able to perform like a seasoned expert.

In other cases, fraud can begin with an authentic, legitimate document, that is then manipulated to misrepresent the truth. These altered genuine documents may originate as real records but are edited to serve fraudulent purposes. For example:

1. Identity theft: One of the most prevalent forms of document fraud, identity theft is the act of obtaining someone’s personal information without their knowledge or consent and using it for fraudulent purposes.
2. Pre-digital document modification: This sophisticated technique involves modifying a digital file, printing it, and then scanning or photographing it to produce a new digital file. Through this process, the altered document loses digital metadata, fingerprints, and traceability that would otherwise reveal tampering. 

In addition, there are other types of document fraud (such as serial fraud, which requires carrying out several of the other fraudulent activities on a mass scale) that have become increasingly common in large organizations.

QUESTION: What types of documents are most commonly faked?

JON KNISLEY: The most commonly faked documents include identification documents (such as driver’s licenses and passports), financial documents (like pay slips and bank statements), and utility bills.

Fraudsters often target these documents because they can provide access to critical services. Criminals will generally use fraudulent documents to misrepresent their identity, financial status, or residency details to give them access to financial benefits and services.

In banking, we see high fraud losses in account origination, credit applications, digital onboarding, and payments. There are always extremely high losses in KYC. 

QUESTION: Most financial institutions already have measures in place to combat fraud, why is it not working?

JON KNISLEY: Indeed, document fraud is unfortunately outpacing traditional controls, and many organizations have incomplete fraud detection solutions. Humans simply cannot keep pace with subtle or AI-generated edits. Some organizations are using siloed tools that create blind spots and fragmented systems that over-flag risk and lack context. Also, global variety is hard to scale – formats, languages, and template drift break traditional checks. Poor-quality inputs (mobile photos, scans, and emails) reduce reliability, and forensic insights often fail to integrate cleanly into core workflows. Another issue is that customer expectations are rising. Clients expect faster onboarding and claims, with zero tolerance for friction. This puts a lot of pressure on teams that are often overloaded with work and are racing to complete onboarding and approvals, providing a weak spot for fraud to slip in.

In addition to these hurdles, limited explainability slows adoption, while audits and regulators expect stronger controls and audit trails as anti-money laundering laws intensify. You only have to look at recent penalties, such as Commerzbank’s $1.55M AML fine to underscore the cost of weak controls.

QUESTION: So what can action can lenders take to protect themselves from document fraud?

JON KNISLEY: In my opinion, it’s vital for lenders to fight back like for like and lean into artificial intelligence. The scale of the problem has led to what is known as Document Forensics, which uses AI to validate the authenticity, integrity, and accuracy of documents and the data they contain. Many lenders currently have incomplete fraud detection solutions because they have implemented simplistic tools that focus only on detecting image manipulation, missing the enormous opportunity that the data contained in the document provides.

Combining checks for image manipulation with rule-based data cross-checks is a more powerful and accurate way to detect fraudulent documents and maintain end-to-end audit trails.

And with the landscape constantly changing, it’s crucial to have solutions that intelligently learn and detect the latest fraudulent techniques, which can be done by using sophisticated technology like machine learning (ML).

However, the most important thing to remember is to have a holistic approach from start to finish. It is crucial to validate documents at the very moment of ingestion and then continue to track risk across all workflows.  Catching fraud requires monitoring the entire process, so you can detect it before it becomes a financial loss.

True document forensics cleans and normalizes poor-quality scans and mobile images, then extracts and structures data from any document worldwide, including different languages and formats. It validates information using rules, cross-checks, and human-in-the-loop controls to detect manipulations such as edited balances, forged fields, metadata tampering, template cloning, and AI-generated artifacts. It also surfaces process-level anomalies like unusual routing or rule violation while reducing false positives that drain operational capacity.

Despite the power of AI tools, I am a true advocate of a human-in-the loop approach as one of the most valuable strategies. This pairs AI speed with human judgment to improve accuracy, safety, and ethical compliance. Experienced employees can identify and correct errors that algorithms miss and provide feedback that allows models to improve over time. Human intervention can also identify biases in training data or AI outputs to ensure fairness and ethical compliance. This will boost transparency and trust and increase the explainability of AI to provide clear explanations for auditors, investigators, and regulators.

QUESTION: What do you think the future holds for document fraud?

JON KNISLEY: We are far from out of the woods in the fight against document fraud. Industry trends suggest it will continue to accelerate. As more financial processes move online, document fraud will grow in both volume and sophistication as AI dramatically lowers the cost and skill required to create convincing fake documents at scale.

Even organizations with strong controls must evolve from basic document verification to full document forensics, cross-validation, and behavioral analysis to keep up with AI-driven document fraud. At the forefront of this is strengthening human-in-the-loop controls. I strongly believe that these strategic measures will ensure that your organization won’t be caught out by a modern-day Frank Abagnale! 

INSIDER PROFILE

Jon Knisley is Head of AI Enablement and Value at global intelligent automation company ABBYY. He works with leading companies to improve their business processes and gain operational insights from critical workflows.