Amid recent technological advances in artificial intelligence and machine learning, on January 17, 2025, the CFPB issued its Winter 2025 Supervisory Highlights: Advanced Technologies Special Edition. This edition of Supervisory Highlights delivers critical industry reminders regarding the balance between regulatory requirements and technological innovation. As an appropriate summation of the CFPB’s overarching worldview, the opening sentence of the Supervisory Highlights explains that “[t]here is no ‘advanced technology’ exception to Federal consumer financial laws.”
In the Supervisory Highlights, the CFPB highlighted instances where credit scoring models used by credit card lenders and auto lenders may result in violations of the Equal Credit Opportunity Act (ECOA) and its implementing Regulation B. For instance, recent CFPB examinations identified disparities in applicant outcomes resulting from the use of credit scoring models in underwriting and pricing credit card applications. The CFPB found disproportionately negative outcomes for protected groups across multiple card products, and critically, examiners suggested that the development or implementation protocols of credit scoring models contributed to the disparities.
According to the Supervisory Highlights, to challenge a disparate impact claim, a financial institution must establish a legitimate business need for a neutral policy or practice that has an adverse impact on a member of a protected class that cannot reasonably be achieved by means that are less disparate in their impact (see12 CFR Part 1002 Supp. I Sec. 1002.6(a)-2). Here, CFPB analysts identified potential alternative credit scoring models that meaningfully reduced disparities while maintaining comparable predictive performance, suggesting that there may be appropriate and less discriminatory alternative credit scoring models that would meet an institutions’ legitimate business needs.
The CFPB’s examiners also noted that financial institutions failed to have adequate compliance management systems (CMS) capable of identifying and addressing these types of fair lending risks. To address these concerns, examiners directed institutions to develop enhanced testing protocols to identify less discriminatory alternative credit models. Examiners required institutions to not only test their credit scoring models but, in the event that testing revealed prohibited basis disparities, to document the specific business needs their credit scoring models serve.
Additionally, in a continuation of a multi-year trend in its messaging, the CFPB also reminded institutions that using “black box” algorithms does not exempt them from providing an applicant with a statement of specific reason(s) for an adverse action as required under ECOA and Regulation B. Examiners found that certain institutions did not sufficiently ensure compliance with adverse action notice requirements and directed the institutions to test the methodologies used to identify principal reasons in adverse action notices.
This special edition of Supervisory Highlights underscores the need for the industry to balance technological innovation with robust compliance frameworks — keeping in mind the impact of any technological advances on existing fair lending laws. To navigate the regulatory landscape, financial institutions should regularly assess their use of artificial intelligence and machine learning models to ensure compliance with applicable laws, including ECOA and Regulation B, and should perform adequate testing to ensure ongoing compliance.