The following post is part of a series of blogs written by MeridianLink® Partners who will be attending MeridianLink LIVE! To learn more about the event, click here.
For as long as financial institutions have been operating, there have been individuals looking to defraud them.
In one study, 98 percent of respondents said they experienced a fraud attempt in 2023. Based on those numbers, it’s not a matter of “if” but “when” a fraud attempt will happen. As the banking industry modernizes, so do the tactics of fraudsters. Generative AI (GenAI) has enabled fraudsters to steal millions of dollars in what is referred to as application fraud (or origination fraud).
Application fraud occurs when false or stolen information is used to apply for a loan or credit. Not long ago, financial institutions could root out fraudsters as they visited a branch during regular business. Now, with online loan applications, FIs must be on the lookout 24/7 for fraudsters operating worldwide.
Some fraud schemes start immediately upon application approval, like straight-roller fraud, when the first 30-60-90 days go by without a single payment. Others, like bust-out fraud, are playing the long game. They establish normal spending and repayment behavior, then abruptly max out their credit limit.
Banks and credit unions face challenges like these and more, but the technology that enables application fraud can also be used to combat it.
Challenges of facing off with application fraud
There are multiple ways application fraud can present. Three of the most common are first-party, third-party, and misreported income.
- First-party: A borrower takes out a loan with no intention to repay it.
- Third-party: A person’s identity has been stolen, or a synthetic identity is created with real and fake information and then used to apply for a loan.
- Misreported Income: An applicant inflates their income on a loan application.
GenAI tools
The simplicity and accessibility of GenAI tools give fraudsters an advantage over lenders. In just a few minutes, they can download tools for voice cloning, falsifying documents, and creating deepfakes for synthetic IDs. Once these materials are created, it only takes a few keystrokes to find a target. Spotting fake information can be difficult, if not impossible, without knowing what to look for.
Increasing roadblocks
A knee-jerk reaction to catch more fraud is to slow down the application process. It’s easy to justify an extra 30 minutes reviewing applications if it prevents thousands of dollars in losses. While you may catch some bad actors with manual reviews, you lose lending efficiency and create a poor consumer experience. Each manual review undermines any underwriting automation you have, and honest consumers who expect speed and convenience may take their business elsewhere.
Limited visibility and reliability
Many third-party solutions provide fraud alerts, but the results don’t always instill confidence in their accuracy or reliability. Most often, the results from popular detection solutions are binary, with few insights into the reasons for the output. When more information is provided, there may only be a single view into limited types of fraud. This doesn’t create a holistic picture of what activities and behaviors should be flagged and why. Inaccurate fraud reporting, like false positives, can also hurt the consumer experience when no fraud has been committed.
Staying ahead with AI
As GenAI technology is exploited in schemes like application fraud, the methods of preventing it are getting more sophisticated. Not surprisingly, the best answer to combating the misuse of AI is AI. Michael Hsu, former Comptroller of the Currency, shared this sentiment in a recent statement, “To successfully fight AI-driven fraud will likely require AI-driven solutions. Banks and AI companies are best positioned to deliver those.”
Machine learning has become a powerful tool for detecting and predicting fraudulent behavior patterns. With fraudsters using deepfakes, stolen identities, and ample targets, relying upon manual reviews and surface-level fraud alerts won’t cut it.
It all comes down to more data and better math to help lenders stay a few steps ahead of fraud. Machine learning models can process thousands of data points to produce accurate results with virtually no disruptions to automated-decisioning. It’s a world where borrowers can still enjoy a convenient and efficient experience, but more fraudsters are stopped before any damage is done.
Smarter, holistic fraud detection with Zest Protect
Zest AI’s fraud detection solution, Zest Protect, helps credit unions and banks stay one step ahead of fraudsters. This advanced, flexible detection tool can seamlessly identify multiple types of application fraud without creating friction for your applicants.
