Auto Buying Fraud is Exploding. Capital One Is Using AI to Fight Back
By Steve Cocheo, Senior Executive Editor at The Financial Brand
Simple Subscribe
Subscribe Now!
Executive Summary
- Auto lending and purchasing fraud keeps rising, hitting $9.2 billion in 2024.
- Fraud targets both dealers and lenders, using human cheating and AI-assisted scams.
- Working with auto dealers Capital One has come up with a new tool — ProtectID — that it says can flag potential fraud by analyzing patterns and connections across multiple forms of consumer credit.
To all appearances, the consumer walking into the auto dealership looked like a hot prospect. He was well-dressed, and after looking over the dealer’s inventory, got excited about a beautiful red Corvette.
Then the dickering began, partially focused on the down payment. The prospect accepted that some kind of down payment is required, but haggled over how much. This is normal behavior, so no cause for concern.
Financing was arranged, and, in the end, the prospect became a customer, and a borrower, and drove off in the $90,000 car.
That’s the end of the story.
The customer/borrower never made a payment and vanished off the face of the earth. To the dealer and the lender, he’s the man who never was.
The dealership and the lender have been victimized by a fraudster, quite possibly one that is connected to a crime ring that makes a business out of stealing cars. Stolen vehicles may wind up being leased through nontraditional channels, others may be shipped out of country for sale with doctored vehicle identification numbers.
Sanjiv Yajnik, president of financial services at Capital One, oversees the company’s auto lending and dealer operation and told that story to a gathering of reporters as an illustration of a growing trend of fraud.
This mushrooming fraud is eating into the already thinning margins in the retail dealership business. Dylan Bullock, general sales manager at Grapevine Dodge, a Texas dealership, said at the event that one of Grapevine’s affiliate dealerships got hit in much this way earlier this year.
The financial loss is hard on the dealership, but Bullock points out that often the fraudsters also often use someone’s legitimate identity to give the appearance of having good credit and other borrowing positives. That only spreads the pain.
Such situations don’t always begin with a walk-in, either. “Many more of our customers are shopping for cars online, post-Covid,” says Bullock, “and we’ve seen a striking increase in fraud there.”
Bullock’s dealership worked with Capital One’s tech staff to develop an AI-based tool called ProtectID to combat auto buying fraud. The banker and the auto executive spoke during the unveiling of the new service that the bank is providing to the industry as part of its Dealer Navigator service.
Read more: The Frictionless Fiction: Why Banking CX Requires Intelligent Pauses
Auto Fraud Comes in Many Flavors, All of Them Bitter
Auto lending/purchasing fraud takes multiple forms today. In 2024, according to an annual study conducted by Point Predictive, auto lending fraud hit an estimated $9.2 billion, which is an increase of 16.5% over 2023.
The leading category of fraud is faked income and employment, the method in 42% of the crimes in 2024. The second-highest category — 27% — is use of synthetic ID and “credit washing.” The former includes the use of “CPNs,” credit privacy numbers, a fraudulent substitute for Social Security numbers often peddled by credit repair scammers who claim the numbers can give a borrower with bad credit a fresh start. Credit washing, on the other hand, is the practice of fraudulently complaining to credit bureaus about entries in one’s files to steadily erase genuine bad items from the record.
But Point Predictive also found that the third-largest (18%) cause of auto fraud is identity theft. This category rose by 29% in 2024. (Point Predictive provides AI analysis of potential frauds in concert with a consortium of over 650 bank and fintech lenders and auto dealer groups.)
“Criminals are increasingly targeting auto loans using stolen identities, adding a more sinister dimension to the fraud landscape that was previously dominated by first party misrepresentations like income inflation and synthetic identity schemes,” says Point Predictive in its report. The report notes that artificial intelligence is enabling many forms of auto fraud. LINKext
Also growing is “bust out fraud.” In this methodology, fraudsters take the time to use synthetic ID or their own identities to build a credit record, making timely payments and otherwise playing by the rules. Once a legitimate record is established, they max out all available credit, including auto loans, and “bust out,” essentially disappearing with prizes like red Corvettes.
Read more: The Hard Truth? Banks’ Lousy Data Management Enables Criminals
Bringing AI to Bear on a Major Challenge to Dealers and Lenders
ProtectID took about a year to develop and is being rolled out gradually to the dealers that work with Capital One, according to Yajnik.
The intent was to use AI to gather and analyze both external data sources such as credit bureau records as well as internal sources at Capital One and among its dealers. Yajnik explains that this includes not only information from the bank’s auto lending activities but also other forms of consumer borrowing, such as credit cards, as well as business credit. This helps connect fraud trends and potential fraudsters across loan types and across multiple institutions.
“Capital One has over 100 million customers in the overall scheme of our institution,” says Yajnik. He says the AI approaches used in ProtectID attempt to make connections across data sources to see commonalities that human analysis could miss.
“We can have a fair amount of certainty about whether someone is bona fide and even if they appear to be connected to someone who is not quite on the right side of things,” says Yajnik. He says ProtectID also attempts to screen out false positives and false negatives found in its preliminary analyses.
Yajnik adds that the sheer volume of the bank’s lending gives it a deep database to draw on, much more than any one dealer group could work with on its own.
The service is built into Dealer Navigator and runs in the background until something is flagged. It is not necessary for the vehicle to be financed by Capital One for the dealer to use the service. Yajnik explains that and Dealer Navigator itself as products the bank supplies to assist dealer businesses without requiring the financing business nor charging fees.
“You have to have the dealer customer’s back if you want to have any shot at having a product that is there for the long terms, because anything that is based on forcing people to do something is short-lived,” says Yajnik.
Read more:
- Capital One Deploys Agentic AI to Support Stressed Auto Dealers (and Finance More Cars)
- Inside Capital One’s Auto Lending Innovation Machine
How ProtectID Works
Here’s the basic process for ProtectID.
When the dealer submits a car loan application to lenders through the system it uses, ProtectID initiates a “soft pull” on the applicant’s credit records and evaluates other factors through the connections made by the AI. Dealers are alerted to potential fraud indicators through the Dealer Navigator system, emails and texts.
The triple notification streams arose from the request of dealer Bullock, who pointed out that dealer staffs are very busy, and may miss messages sent in only one channel.
Having identified potential risks, the system suggests steps for addressing them, such as ways to verify identity. Additional information supplied by the applicant, from phone number validation to selfies, are reviewed by Capital One fraud detection processes.
Yajnik says as important as it is to stop fraudsters, legitimate customers who hit a ProtectID tripwire must have the opportunity to show that they are bona fide.
He adds that ProtectID is at a beginning, not a conclusion.
“Tuning is very important and it’s not a one-day thing,” says Yajnik. “It’s complex, so you don’t end up with false positives and false negatives. You have to keep tuning, tuning, tuning. It’s a never-ending issue.”
Identifying potential fraud factors and pointing out possible remedies leaves room for the human factor, according to Dylan Bullock.
“We’re going to do the same thing we’ve always done,” says Bullock. “We’ll assess the situation, assess the deal, and then make a decision. Because ultimately it is still our decision as the dealer.”
Read more about fighting fraud:
