CFPB Addresses Adverse Action Notices Resulting from AI Credit Models

The CFPB recently published Circular 2023-03, which states that creditors using artificial intelligence (AI) or complex credit models must provide adverse action notices with specific and accurate reasons for taking the adverse action in order to comply with ECOA.

The Circular reiterates the CFPB’s position that ECOA, including the adverse action notice requirement, applies no differently to credit decisions made using AI models or “black box” algorithms.  Under ECOA, it is unlawful for creditors to discriminate against credit applicants on the basis of a protected characteristic, and a creditor that takes an adverse action against an applicant must provide the applicant notice of that action and specific reason(s) the action was taken.  The CFPB notes that although Regulation B provides sample notice forms, the reasons provided on the sample forms should not be used by creditors if they do not accurately reflect the principal reason an adverse action was taken.  The CFPB further states that merely selecting the “closest” reason on the sample form is not compliant with ECOA if the reason is inaccurate.

The Circular explains that creditors who use AI or other complex algorithms to make credit decisions might rely on consumer surveillance data or other data not included in an applicant’s credit file or application when making a credit decision.  Because consumers may not be aware that additional information may be used to reach a credit decision, the CFPB believes specificity in adverse action notices is of particular importance when an AI model is used to reach a credit decision.