Weekly Fintech Focus

  • The CFPB terminates a no-action letter with an AI credit underwriter.
  • A CFPB circular confirms that AI underwriting models are subject to anti-discrimination laws including adverse action notices.
  • BNPL companies and credit bureaus face urging by the CFPB to properly report consumer information.
  • The CFPB launches an initiative to improve customer service at big banks.

The CFPB Terminates AI Credit Underwriting No-Action Letter

On June 8, 2022, the Consumer Financial Protection Bureau (CFPB) issued an order terminating a no-action letter (NAL) that the CFPB had originally granted to lending platform Upstart Network, Inc. (Upstart), in 2017 (the CFPB’s first-ever NAL) and subsequently renewed in November 2020 for a three-year term.

Under the CFPB’s NAL Policy, a person can apply for no-action treatment on a new product or service that offers the potential for significant consumer benefit where there is uncertainty about how the CFPB would apply specific provisions of law. A grant of no-action treatment signifies that the CFPB has no present intent to initiate supervisory or enforcement action against the recipient with respect to the subject of the NAL.

The CFPB’s termination of the Upstart NAL is the latest in a series of actions by the CFPB that have raised questions about the future of its NAL Policy. The following timeline briefly summarizes these events:

  • In 2017, Upstart requested a NAL from the CFPB to clarify that Upstart’s credit underwriting model, which involved proprietary applications of artificial intelligence and machine learning to supplement traditional credit scoring methodologies, did not present a violation of the Equal Credit Opportunity Act (ECOA) and Regulation B. The CFPB granted Upstart’s request, making Upstart the first entity to receive no-action treatment under the CFPB’s then-new NAL Policy.
  • On November 30, 2020, the CFPB renewed Upstart’s NAL for another three years. The terms and conditions of the NAL required Upstart to notify the CFPB of significant changes to Upstart’s model.
  • On April 13, 2022, Upstart notified the CFPB that it intended to add new variables to its underwriting and pricing model. According to the CFPB’s June 8 termination order, CFPB staff requested more time to review Upstart’s proposal.
  • On May 24, 2022, the CFPB announced that it was replacing its Office of Innovation (which processed the applications for NALs) and its Project Catalyst (another initiative designed to encourage innovation) with a new Office of Competition and Innovation. The CFPB’s press release stated that “[a]fter a review of these programs, the agency concludes that the initiatives proved to be ineffective and that some firms participating in these programs made public statements indicating that the Bureau had conferred benefits upon them that the Bureau expressly did not.”
  • On May 27, 2022, according to the CFPB’s termination order, Upstart requested that the CFPB modify the NAL to reduce its term from 36 months to 18 months, meaning that it would terminate three days later on May 30, 2022.
  • On June 8, 2022, the CFPB announced that it had issued the order to terminate Upstart’s inclusion on its list of approved NALs.

A CFPB Circular Confirms Anti-Discrimination Laws Apply to Algorithms

The CFPB has issued a circular confirming that federal anti-discrimination law requires explanations of specific reasons for denying credit applications or taking other adverse actions against applicants. The circular warns companies using algorithmic decision engines (or AI engines) that a “black-box model” for lending decisions does not absolve the company from explaining adverse actions to applicants as required by law. The agency warns that with some black-box models, users and developers may not be able to know the reasoning behind the model’s outputs, which could result in companies being unable to meet the ECOA’s adverse action notice requirements. The ECOA and Regulation B require a creditor to provide notice when it takes an adverse action against an applicant, explaining with specific and accurate reasons why the creditor took such action. If the creditor is using technology that does not enable the creditor to explain its decision-making process, then the creditor will not be able to comply with law. In short, complexity, opacity, or time in the market will not be considered excuses for failure to meet a creditor’s adverse action notice requirements.

BNPL Companies Urged by CFPB to Report Credit Data

On June 15, 2022, the CFPB published a blog post following up on its inquiry (which we discussed here) into “buy-now pay-later” (BNPL) companies. In the post, the agency urges BNPL companies to report positive and negative data to credit bureaus when BNPL payments are furnished. Further, the CFPB encourages the BNPL industry to develop standardized BNPL furnishing codes and formats to provide data that fits with the unique BNPL product offering. Although the major credit bureaus have announced plans to accept BNPL data, the CFPB is concerned that the differences between the credit bureaus’ plans will result in inconsistent treatment of such data, meaning the furnishment of such data will have less benefit to consumers. The CFPB will monitor the BNPL industry’s progress as changes to the reporting of consumer credit data are implemented.

The CFPB Launches Initiative to Improve Customer Service at Big Banks

CFPB Director Rohit Chopra led a town hall meeting on June 14, 2022, in Great Falls, Montana, to discuss the agency’s new initiative. The CFPB is requesting comments from consumers about their relationship with their banks, including how they assert their rights to better service with big banks and credit institutions. The town hall included local community organizations, advocates, leaders, and members of the public. Together, the group discussed the challenges faced by rural Montanans and how banking deserts adversely affect Montana’s financial landscape.

Chopra noted that recent bank consolidation has had mixed results for consumers and customer service experiences, particularly in rural communities. Rural customers faced decreased banking access as they were more likely to visit smaller banks or credit unions but are now living in rural banking deserts with no intimate relationship banking.

Additionally, many financial institutions and tech companies are shifting toward what Chopra calls “algorithmic banking,” which relies on using vast quantities of data about an individual through tracking and surveillance to make predictions about their behavior and banking habits. Chopra accepts that the shift away from traditional relationship banking could eliminate discrimination based on human judgment, but cautions that automated technologies also pose a concern as algorithmic bias can affect outcomes in an unjust way.

To reinvigorate relationship banking, the CFPB has issued a Request for Information to find out how people can assert their rights to better customer service with their depository institution. “Customers of large banks should not have to run through an obstacle course to get a straight answer about their account,” said Director Chopra during the town hall meeting. Notably, the CFPB is looking to understand, among other things, (i) the types of information people request from their bank and how they are using that information; (ii) what information is currently unavailable to consumers from their banks; and (iii) any customer service obstacles that inhibit a consumer’s ability to bank (e.g., wait times, disconnected calls, or the quality of responses to questions).

The CFPB also seeks to guarantee that algorithmic banking does not receive preferential treatment and must adhere to the same laws as traditional banks. The CFPB published a policy in March confirming that financial companies must explain to applicants the specific reasons for denying a credit application or taking other adverse actions. It has also ordered several Big Tech companies, such as Facebook, Apple, and Google, to provide the CFPB with information on their efforts to gain more control over payment systems and how they plan to use customer data to feed their algorithms.