Are Borrowers Entitled to an Explanation When AI Denies Their Loan Application?

The Legal Right to Know Why

In the United States, the Equal Credit Opportunity Act, along with its implementing regulation, requires lenders to provide applicants with specific reasons when a credit application is denied, or when a lender takes other “adverse actions” like offering less favorable terms than requested. This disclosure, known as an adverse action notice, has to identify the actual principal reasons behind the decision rather than offering only a vague or generic statement. This legal requirement predates modern AI-driven underwriting, but regulators have made clear it applies fully to AI-based credit decisions as well.

Why This Becomes Complicated With Complex AI Models

The challenge with more sophisticated machine learning models is that they can combine many variables in complex, non-linear ways that make it harder to point to a small number of clear, individually meaningful reasons for any specific decision — the “black box” problem mentioned frequently in discussions of AI transparency. A traditional, simpler scoring model might make it relatively easy to say “your application was denied primarily because of your debt-to-income ratio and length of credit history.” A highly complex model might arrive at a similar decision through a much less straightforward combination of factors.

Federal regulators, including the Consumer Financial Protection Bureau, have directly addressed this issue, stating that lenders cannot use the technical complexity of their models as a justification for failing to provide the specific, legally required reasons for a denial. In practice, this pushes lenders toward using explainability techniques that can identify which factors most influenced a given model’s output, or toward choosing models that are inherently more interpretable, specifically so they can meet these disclosure obligations.

What a Compliant Explanation Actually Looks Like

An adequate adverse action notice generally needs to identify the specific, primary factors that led to the negative decision, expressed in terms the applicant can understand, such as “insufficient length of credit history” or “high ratio of debt to available credit,” rather than only a numeric score or a vague reference to “overall risk assessment.” Lenders are expected to be able to trace a denial back to genuine, identifiable factors in the applicant’s data, and providing only a bare model output score without meaningful reasons generally would not satisfy the legal requirement.

Bottom Line

Borrowers are legally entitled to specific reasons when a loan application is denied, and this right applies to AI-driven credit decisions just as it does to traditional underwriting; regulators have made clear that a lender’s use of a complex or opaque AI model doesn’t excuse it from providing a real, understandable explanation for the denial.

Go deeper

• What Is a ‘Black Box’ AI Model?