On September 19, 2023, the Consumer Financial Protection Bureau (CFPB) issued new guidance outlining lenders’ obligations under the Equal Credit Opportunity Act (ECOA) and Regulation B when using artificial intelligence (AI) or advanced algorithms in credit decisions. The crux of the guidance is clear: lenders must go beyond merely using AI; they must also understand how it learns, processes data, and ultimately makes decisions.
This move by the CFPB emphasizes that lenders may not simply rely on the CFPB’s own model forms for adverse action notifications, especially when intricate algorithms or AI come into play.
While these traditional forms, with their standard reasons for credit denial such as “insufficient income” or “poor credit history,” have been foundational, the financial landscape is evolving. And if your algorithms analyze unconventional data like a consumer’s online habits or geographic patterns—or if you’re uncertain about which data points they assess—there’s a significant likelihood that the standard forms will fall short. Simply put, if AI or algorithms are crucial to your credit evaluations, their logic must be as transparent as a human’s.
Consider an underwriting algorithm that bases a denial of credit on a consumer’s occupation. In such scenarios, broad justifications like “insufficient income” become inadequate. Even if the core reason aligns with familiar concepts of income or credit standing, the CFPB demands more specificity. Lenders must pinpoint and convey the exact principal reason guiding the decision.
This new era of credit decision-making brings with it a set of unique challenges. Not only do AI models often rely on non-traditional data points that might be obscure to the average consumer, but the inherent “black box” nature of certain AI systems can also make their reasoning elusive even to the institutions deploying them. But the CFPB insists that unfamiliarity is not an excuse. Lenders have an obligation to ensure that their adverse action notices specifically and correctly outline the factors that influenced the credit decision, even if those factors are derived from advanced algorithms.
In response to this guidance, lenders should initiate a comprehensive review of their AI models and complex algorithms to ensure compliance with ECOA and Regulation B. Institutions utilizing AI or algorithms should focus on the “explainability” of these models. Beyond validating the model’s performance, they must ensure that clear, understandable reasons for credit decisions can be extracted. This may involve investing in AI transparency solutions or consulting with experts who can decipher complex algorithms.
As AI continues to play an increasingly significant role in modern financial services, it’s crucial for lenders to balance cutting-edge technologies with strict and ever-evolving regulatory standards. It’s a challenge, but one that must be met head-on.
What this means to you
Re-evaluate your AI tools. If you're employing AI or advanced algorithms for credit decisions, it's pivotal to understand not just their outputs, but also their inner workings. This means a deeper dive into how they process and interpret data.
CFPB’s model forms may fall short. Traditional adverse action notification forms might not cut it anymore, especially when AI models consider unconventional data. It's crucial to pinpoint and provide specific reasons for credit decisions.
Understanding over unfamiliarity. The inherent “black box” nature of some AI systems doesn’t exempt you from explaining credit decisions. Even if an AI model’s reasoning seems elusive, the CFPB insists on clarity in adverse action notices.
A call for review and compliance. Ensure you’re up to date with the ECOA and Regulation B standards, especially in the context of AI. As AI becomes more prevalent in financial services, striking the right balance between tech innovation and regulatory standards is crucial.
[View source.]
