Using Biometrics and AI Computer Vision to Authenticate Borrower Identity

How do you know your borrower is who they say they are? Do you rely on a photo of their driver’s license only? This could be a fake or manipulated photo. Do you do a live video call with them to verify the person taking out the loan matches the IDs that you received?

In an unofficial poll I conducted on LinkedIn, 93% of mortgage and title respondents said they were NOT using biometric data to verify borrower identities. According to CoreLogic:

“We estimate that in the second quarter of 2023, 0.75% of all mortgage applications were estimated to contain fraud, about 1 in 134 applications. The highest risk segment remains 2- to 4- unit properties, with 1 in 28 transactions estimated to have indications of fraud.”

These estimates are from a down cycle time in the mortgage industry, so fraud is potentially higher during more prosperous times. This doesn’t detail the exact type of fraud committed, but if more due diligence was put into verifying borrower’s identities if could stifle any and all fraud on the loan. Imagine if lenders recorded a video of the borrowers upfront and explained that these videos are now packaged with the loan to prevent fraud. That could put a serious damper on scandalous activity.

Once a lender has video of all borrowers, they could run AI computer vision and AI facial recognition software to match the faces in the video to the picture on the IDs.  They can verify that the measurements of eyes, nose, mouth, etc match up on both.

Lenders could say that the real estate agent or title company sees the borrower in person and can handle the verification. I assure you that real estate agents probably do not check IDs and if they do they aren’t trained to authenticate the ID. Neither are title companies. And verifying borrower identity at the closing table is way too late in the process. Also, the real estate company and the title company do not have as much on the line as the lender does.

I’m betting that the GSE’s will eventually require higher standards of borrower identification in the near future.