The Consumer Financial Protection Bureau is putting artificial intelligence-based (AI) underwriting models under the microscope as more lenders and banks rely on complex algorithms in the credit-decisioning process.

Specifically, the Bureau wants to ensure that lenders using AI or machine learning (ML) to underwrite loans are able to issue adverse action notices required under the Equal Credit Opportunity Act (ECOA), a CFPB spokesperson told Bank Automation News. ECOA requires that lenders explain in writing why a particular consumer was denied access to credit through an adverse action notice.

“The CFPB doesn’t set standards for AI. Rather, the CFPB interprets and enforces federal consumer financial law, which applies regardless of the technology used,” the spokesperson said, noting that lenders may be liable under the Unfair, Deceptive or Abusive Acts of Practices Act (UDAAP), as well.

The Bureau is concerned that complex algorithms prevent lenders from explaining why a particular consumer was denied credit, which is a direct violation of both ECOA and UDAAP, according to a May 26 CFPB bulletin.

“The Bureau has regularly emphasized the importance of compliance management systems to detect and prevent violations of Federal consumer financial law, including systems specifically designed to manage potential violations arising from the use of algorithms,” the spokesperson said. “CFPB examiners may test the accuracy of adverse action notices and may request documentation to assess legal risks.”

Pumping the brakes

Fair lending and the use of technology has been one of the Bureau’s top priorities since Rohit Chopra took over as director of the CFPB in 2021, Brian Fink, an attorney at McGlinchey, said.

“Director Chopra mentioned in his [Senate] confirmation testimony that AI algorithms, machine learning and fair lending are priorities, and I think this is a twofor,” Fink said. “This nominally addresses the fair lending issue. I think they took the opportunity to essentially throw a bunch of sand in the gears of folks who are using machine learning. It puts the breaks on it, in some way.”

Champions of AI and ML-based underwriting frequently argue that the technology enables access to credit for a segment of consumer that wouldn’t otherwise qualify under traditional, FICO score-based systems. And while that may be true, the Bureau is likely concerned that the proliferation of these new technologies and their ability to learn “creates risks that no one understands,” Fink said.

“That’s the concern from a fair lending perspective — these tools that could broaden the pie…maybe they only do for some groups,” Fink said. “And that’s problematic. I think that’s what they’re doing with this adverse action circular is trying to make companies who use this tech pause and evaluate what the impacts are and the causes of their decisioning and if they understand them fully.”

The move comes on the heels of the CFPB’s December 2021 call for whistleblowers, another signal of the Bureau’s increased interest in AI and fair lending.

“Tips from whistleblowers can inform the Bureau’s work by giving it insight into consumer financial acts or practices that cause consumer injury or otherwise violate the law but might otherwise not be visible to consumers or the Bureau,” the spokesperson said. “This is especially true with respect to discriminatory practices that may be hidden in algorithms or other technologies.”

To be sure, the Bureau has authority to ask lenders how their technology works in the examination process, the spokesperson said, “but whistleblower tips can help the Bureau focus its limited resources on potential wrongdoing.”

The CFPB has been stepping up its visibility in recent months, especially in auto lending. In March, the Bureau acted to halt illegal auto repossessions, and in May, eyed miscalculated ancillary product refunds as a top priority.

–Additional reporting by Felix Behr

Bank Automation Summit Fall 2022, taking place Sept. 19-20 in Seattle, is a crucial event on automation and automation technology in banking. Learn more and register for Bank Automation Summit Fall 2022.