The Consumer Financial Protection Bureau (“CFPB”) released a regulatory Circular providing guidance regarding the use of “complex algorithms” to assess whether a consumer should be extended credit. Often referred to as “black box” solutions, which may include artificial intelligence protocols, the CFPB has stated that full compliance with obligations is required, regardless of the technology used.
One of the challenges that “black box” technologies present is that the reasons for the results received are not provided and cannot be easily ascertained by humans. The CFPB is not forbidding the use of “black boxes” but is effectively mandating the use of so-called “explainable AI” procedures in association with black box solutions. “Explainable AI” basically means that a subsequent black box (or series of black boxes) is added onto the solution, which is trained specifically to identify the reasons the first black box reached its results. Thus, when a creditor uses a black box to assess whether to extend credit to a consumer, secondary black box programs can be utilized to provide the specific reasons that the consumer was not approved, which is required by the Equal Credit Opportunity Act to be identified on adverse action notices that are then provided to the consumer.