CFPB Continues Analysis of Algorithmic Technology | Alston and Bird

On May 26, 2022, the Office of Consumer Financial Protection issued a Consumer Financial Protection Circular stating that creditors using tool algorithms in credit decisions must provide “specific statements of reasons” in accordance with ECOA and B. The CFPB had previously stated that the circular was “a policy statement aimed at providing guidance to other agencies with responsibilities for consumer financial protection to determine how the CFPB intends to comply with federal consumer finance law.” The lawsuit alleges that some complex algorithms are a “black box” that cannot be interpreted, which makes it difficult, if not impossible, to accurately identify specific reasons for denying credit or other harmful actions. The CFPB concluded:[a] the creditor cannot justify the failure to comply with the requirements of the ECOA and Regulation B in understanding that the technology used to evaluate the applications is too complicated or too opaque.

This latest circular follows a CFPB proposal related to the review of AI used in automated valuation models (“AVMs”). As we stated in our previous post on this topic, the CFPB stated that some algorithmic systems may violate the ECOA and the Implementing Regulation (“Regulation B”). In this preliminary scheme of data entry proposals, the CFPB acknowledged that some machine learning algorithms may be “too opaque” for auditing. The CFPB also theorized that algorithmic models could “repeat historical patterns of discrimination or introduce new forms of discrimination because of the way a model is designed, implemented, and used.”

According to Regulation B, the statement of reasons for the adverse action taken “must be accurate and state the main reason (s) for the adverse action. Statements of failure to score are not sufficient. ” In the circular, the CFPB reiterated that when using disclosure form templates, “if the reasons in the forms are not the factors actually used, a creditor will not comply with the notification requirement by verifying the closest identifiable factor listed.” In another related advisory opinion, CFPB- He also confirmed earlier this month that the ECOA and Reg B provisions apply not only to credit applicants but also to those who have already received credit.This attitude echoes the previous amicus summary on the same subject presented by the Bureau. John Fralish v. Bank of Am., NA, zk. 21-2846 (L), 21-2999 (7. Zir.). As a result, the CFPB affirms that the ECOA requires lenders to issue “warnings of adverse action” to lenders with existing credit. For example, the CFPB asserts that the ECOA prohibits lenders from lowering the credit limit on certain loan accounts or subjecting certain loans to more aggressive prohibitive collection practices, such as race.

The latest CFPB circular indicates a less favorable view of AI technology than the previous statements by the Bureau. In a July 2020 blog post, the CFPB highlighted the benefits of using consumer AI or machine learning in credit recruitment, stating that “it has the potential to expand credit access, allowing lenders to assess the credit quality of millions. The CFPB also acknowledged that uncertainties over the current regulatory framework could slow down the adoption of the technology, which the CFPB said at the time maintained the ECOA’s “flexibility” and that “a creditor did not have to describe how or why a known factor affected an application. … or, for credit scoring systems, how the factor relates to credit quality. ”In a previous post, the CFPB concluded: don’t be clear about it. This flexibility can be useful for creditors when issuing notices of harmful actions based on known AI models and key causes, but based on non-intuitive relationships. This publication also highlighted the Office’s Non-Action Letter Policy and Sandbox Support Policy as a tool to help provide an AI development shelter. However, in a recent statement, the CFPB criticized the ineffectiveness of these programs and it appears that these programs are no longer a priority for the Bureau. Also, this previous blog post now contains a disclaimer: “It provides an incomplete description of the ECOA and Regulation B action statement, which apply equally to all credit decisions, regardless of the technology used to make them. ECOA and B regulations do not It allows the creditor to use technologies that cannot give specific reasons for harmful actions.

[View source.]

Leave a Comment