Creditors Using Complicated Algorithms Must Explain Credit Denials

Creditors must give consumers an explanation when denying credit—they can’t just blame a computer algorithm.

By , Attorney

The Equal Credit Opportunity Act (ECOA) (15 U.S.C. § 1691 and following) is one of the primary federal laws prohibiting credit discrimination. The ECOA prohibits creditors from refusing to grant credit because of race or color, national origin, sex, marital status, religion, age, public assistance status, or because the applicant has in good faith exercised any right under certain consumer protection laws.

Under the ECOA, a creditor must give you a notice if it denies you credit or takes another adverse action against you. The notice must give specific and accurate reasons for the adverse action. (An "adverse action" includes denying a credit application, terminating an existing credit account, making unfavorable changes to an existing account's terms, and refusing to increase a credit limit.)

For years, creditors and lenders have used data harvesting (collecting information from browsers, websites, stores, and more) and computational methods when making lending decisions. And for the most part, creditors that used this technique have been able to provide rationales for their credit decisions. But now, some creditors use so-called "black-box" algorithms when making credit decisions. These systems incorporate complicated machine learning and artificial intelligence, so it's difficult, if not impossible, for the creditor to identify the exact reasons the computer system denies an applicant for credit. Sometimes, the reasoning for accepting or denying a loan is not only unknown to the system's users but also to the system's developers.

Nevertheless, the Consumer Financial Protection Bureau (CFPB) recently confirmed that the ECOA requires creditors to provide detailed adverse action notices. This requirement applies whether the creditor uses a sophisticated algorithm or more conventional ways to evaluate an application. So, creditors and lenders can't just claim an algorithm made the decision or say that the method used to assess an application is too complex or difficult to understand. It must disclose the specific principal reasons for denying credit or taking another adverse action.

Effective date: May 26, 2022