In September, the Consumer Financial Protection Bureau issued guidance on compliance with the Equal Credit Opportunity Act’s adverse action notice requirements for lenders utilizing artificial intelligence in their credit decision process. Specifically, the Bureau noted that lenders may not rely solely on the checklist of reasons identified in the model adverse action notice under Regulation B, but instead must provide the specific reasons the lender relied upon to take the action it did.
What You Need to Know:
- Lenders are increasingly utilizing artificial intelligence in credit decisions.
- The CFPB’s guidance warns these lenders that adverse actions taken against consumers must be described with sufficient specificity to provide the consumer with the actual reason for the action taken.
- Use of the checklist on the model form will not be sufficiently specific in the Bureau’s view.
ECOA and its implementing regulation, Regulation B, require that a creditor must provide an applicant for credit with a statement of reasons for any adverse action taken against the applicant, such as a denial of credit. Regulation B includes a model form creditors may use to meet these obligations, and the model form includes a checklist of reasons that may be used. This form may be used to satisfy obligations under both ECOA and the Fair Credit Reporting Act, although these requirements are different.
The Bureau notes that the checklist items are only a list of commonly used reasons for adverse actions, and are not the only reasons why a creditor may take adverse action against an applicant. Specifically, when a creditor is using artificial intelligence and/or other algorithmic or predictive models to assist it in its credit decision, the Bureau believes that the checklist items are insufficient to provide consumers with the reasons for the adverse action. The Bureau notes that creditors’ use of the most similar checklist item will not satisfy the disclosure obligation, and the creditor will be in violation of ECOA.
The Bureau believes that specificity in the adverse action notice is particularly important where AI or other complex algorithms are used, because consumers may not expect that data gathered outside of the application process will be utilized or affect the credit decision. If the decision is based on patterns of behavior, such as the type of business frequented or type of goods purchased, the creditor would need to disclose specific details about the purchase history or patronage that led to the adverse action. The Bureau extends this opinion to adverse actions taken on new applications for credit as well as adverse actions taken on existing lines of credit.
In its press release for this guidance, the Bureau also emphasized that it “has made the intersection of fair lending and technology a priority.” It views this guidance as an additional piece of its enforcement of fair lending laws.
The Bureau’s guidance, in essence, is that adverse actions must be described with sufficient specificity to provide the customer with the actual reason for the adverse action, even if it may upset the consumer. While the Bureau does not identify what would be sufficiently specific to comply with ECOA, it provides examples of what it thinks would be insufficient. Lenders utilizing or considering utilizing algorithmic credit decisioning should keep these requirements and this guidance in mind as part of their assessments.
