Skip to main content
    Kriv AI

    AI in lending

    AI in lending: what it is, what Regulation B says about denials and what a lender should check

    A plain guide for lenders, credit risk and compliance leads. Facts about the CFPB and Microsoft come from their own pages as read on 11 November 2026. This page is not legal advice.

    AI in lending means using machine learning or other automated models in credit decisions or the work around them, such as reviewing documents, scoring applications and flagging risk. Under Regulation B, 12 CFR 1002.9, a creditor that takes adverse action must give the applicant either a statement of specific reasons for the action taken or a disclosure of the right to that statement. If the creditor only discloses the right to the statement, it is due within 30 days when requested within 60 days of the creditor's notification. Read on 11 November 2026. Before a lender uses a model, it should be able to name the specific reasons behind a denial and show who reviewed them.

    plain answer

    What is AI in lending?

    In this guide, AI in lending means any model that learns from data and is used in credit work. That can be a model that scores applications, a tool that reads income documents, or an assistant that drafts letters to applicants. The risk grows with how much the output decides: a draft a person rewrites is different from a score that decides a loan.

    Microsoft's Azure Machine Learning documentation, read on 11 November 2026, uses lending as an example of fairness: AI systems that give guidance on loan applications should make the same recommendations to people with similar financial circumstances or qualifications.

    Sources: Microsoft Learn, What is Responsible AI (https://learn.microsoft.com/en-us/azure/machine-learning/concept-responsible-ai), read 11 November 2026.

    plain answer

    AI lending and the CFPB: do you still have to explain a denial?

    Regulation B implements the Equal Credit Opportunity Act. Its section 1002.9 covers notices. The description below is the general rule; 1002.9(a)(3) sets different rules for business credit. Under 12 CFR 1002.9, a notification given to an applicant when adverse action is taken must be in writing and must contain, among other required items, either "a statement of specific reasons for the action taken" or a disclosure of the applicant's right to that statement. The creditor must then give the statement within 30 days if the applicant asks for it within 60 days of the creditor's notification. The CFPB withdrew its Circular 2022-03 on adverse action and complex algorithms in May 2025, so this page does not rely on it.

    This page does not say whether any particular lender or model meets that rule. Your counsel decides what applies to you.

    Sources: eCFR, 12 CFR 1002.9 Notifications (https://www.ecfr.gov/current/title-12/chapter-X/part-1002/section-1002.9), read 11 November 2026. Federal Register 2025-08286, Interpretive Rules, Policy Statements, and Advisory Opinions; Withdrawal, 12 May 2025 (https://www.federalregister.gov/documents/2025/05/12/2025-08286/interpretive-rules-policy-statements-and-advisory-opinions-withdrawal), read 11 November 2026.

    how to

    What should a lender check before using AI in credit decisions?

    This list is our own view, not the CFPB's. Ask whether you can name the specific reasons behind each denial in terms an applicant can read. Ask who reviews a model's output and when a person can overrule it. Ask whether results have been checked across different groups of applicants, in line with the fairness concern Microsoft's page describes. Ask what record is kept of each score, each reason and each decision. Ask who can switch the model off, and who owns it after the vendor or the builder leaves.

    If a vendor supplies the model, ask what they can show you about how it reaches its reasons.

    kriv role

    Where Kriv helps

    Kriv AI helps lenders design the review steps, records and testing around AI, and train the teams who run them. Kriv AI has published articles on fair lending explainability and bias monitoring for credit models and on model risk management for mid-market lenders, linked below. Kriv AI does not make credit decisions, sell a scoring model or give legal advice, and nothing here is a statement about any specific system. Scope and price are set after a short discovery call.

    limits

    What this page could not confirm

    We read Regulation B, 12 CFR 1002.9, on eCFR, the Federal Register notice that withdrew CFPB guidance in May 2025, and one Microsoft page. We did not read the rest of Regulation B, its official commentary, the Fair Credit Reporting Act, state rules or any regulator's current position. The only figures we quote are the 30 day and 60 day periods in the rule. For anything not cited above, check the CFPB's current pages and ask your counsel.

    Straight answers

    Frequently asked questions about AI in lending: what it is, what Regulation B says about denials and what a lender should check

    What is AI in lending?

    Using machine learning or other automated models in credit decisions or the work around them, such as document review, scoring and risk flags. This is our plain definition for this guide.

    Can a lender use a complex AI model and still deny credit?

    Regulation B requires a creditor that takes adverse action to give the applicant a statement of specific reasons for the action taken, or a disclosure of the right to that statement (12 CFR 1002.9, read 11 November 2026). Ask your counsel how it applies to you.

    How long does a lender have to give the reasons for a denial?

    Under 12 CFR 1002.9, if the notice only discloses the right to a statement of specific reasons, the creditor must provide it within 30 days, if the applicant requests it within 60 days of the notification (eCFR, read 11 November 2026).

    Should a person review AI credit decisions?

    We recommend that a person can review and overrule a model's output on anything that affects an applicant. This is our design view, not a rule we are citing.

    Does Kriv AI make credit decisions or sell a scoring model?

    No. Kriv AI designs the review steps, records and testing around AI and trains teams. It does not give legal advice.

    Talk to the team that would do the work

    Bring your requirements to a working session with the person who'll actually deliver.

    Book a Discovery Call