On January 17, the CFPB released a report titled “Supervisory Highlights: Advanced Technologies Special Edition” which highlights the examination of certain lenders that utilize AI and machine learning when making credit decisions and their compliance with ECOA and Regulation B. As it relates to credit card lenders, the report found disparities in outcomes for Black or African American and Hispanic applicants compared to white applicants, despite institutions using AI and machine learning in credit decision-making. The CFPB suggested that the way the institutions developed or implemented their credit scoring models may have contributed to these disparities. To facilitate compliance with ECOA and Regulation B, the CFPB examiners recommended that the institutions “develop a process for considering a range of less discriminatory models when a model produces prohibited basis disparities (for example, alternative models generated through automated testing).”
The CFPB also examined auto lenders, where credit scoring models with numerous input variables, like “alternative data,” posed significant consumer protection risks. Examiners found institutions lacked processes for reviewing input variables for fair lending risks and did not justify different model inputs contributing to these disparities. Additionally, the report addressed adverse action notices, emphasizing creditors must provide reasons for adverse actions, even when using an algorithm.
The CFPB notes that even when using AI and machine learning, financial institutions are still under the obligation to comply with federal consumer financial laws.