The Consumer Financial Protection Bureau has doubled down on adverse action requirements for lenders using AI-based underwriting with new guidance, marking a shift in tone from its previous guidance in 2022.  

A CFPB blog post Tuesday echoed guidance issued in 2022, but largely focused on adverse action notices’ accuracy and specificity requirements, Brian Fink, partner at McGlinchey, told Auto Finance News, a sister publication to Bank Automation News,

The Equal Credit Opportunity Act requires lenders to send adverse action notices to potential borrowers about why they were denied credit. The bureau has voiced concerns that some creditors using “black-box” underwriting algorithms may not be able to explain the algorithm’s decisions. 

The 2023 and 2022 guidance are “remarkably similar,” Fink said. “This guidance focuses on accuracy and specificity requirements. They’re saying … you need to drill down on the specific reason — not a close reason. It has to be accurate.”  

Notably, the CFPB’s most recent guidance marks a shift in the regulator’s tone, Fink said.  

“In the first [guidance] they said that they think people may not understand the models that they use,” he said. “In the second one in 2023, they sort of take the tack of: People may be doing this on purpose.”  

The bureau’s Sept. 19 Consumer Financial Protection Circular provides an example of the sort of specificity it seeks in adverse action notices,” Fink said.  

“For instance, if a complex algorithm results in a denial of a credit application due to an applicant’s chosen profession, a statement that the applicant had ‘insufficient projected income’ or ‘income insufficient for amount of credit requested’ would likely fail to meet the creditor’s legal obligations,” according to the circular.  

“Even if the creditor believed that the reason for the adverse action was broadly related to future income or earning potential, providing such a reason likely would not satisfy its duty to provide the specific reason(s) for adverse action.”  

Any consumer creditor, including auto lenders that use AI algorithms for underwriting is required to comply with the CFPB’s guidance, Fink said.  

Editor’s note: This article first appeared on Auto Finance News, a sister publication to Bank Automation News.