AI in financial services
AI credit scoring: what it is and what a lender has to explain
A plain guide for lending, risk and compliance leads. The legal statement comes from Regulation B as read on 11 November 2026. The definitions and the checklist are our wording.
AI credit scoring is the use of machine learning models to help estimate how likely an applicant is to repay, and so to support a credit decision. It does not replace the lender's responsibilities. The Equal Credit Opportunity Act and Regulation B require a lender to give an applicant a statement of specific reasons, or tell the applicant how to ask for one, when it takes adverse action (12 CFR 1002.9). That means a lender using an AI model still has to be able to give an applicant the specific reasons for a refusal. This page covers what to record and who should own it. Source read 11 November 2026.
plain answer
AI credit scoring: what is it?
In plain terms, AI credit scoring means a model learns patterns from past lending data and produces a score or recommendation that a person or a rule then uses to approve, decline or price a credit request. That is our wording, not a legal definition.
The model is one input to a decision. The lender still owns the decision, the policy behind it and the notice the applicant receives.
plain answer
Do lenders still have to give reasons when an AI model decides?
Yes. The rule that stands is Regulation B, 12 CFR 1002.9. When a creditor takes adverse action, the notice to the applicant must contain either "a statement of specific reasons for the action taken" or a disclosure of the applicant's right to a statement of specific reasons within 30 days, if the statement is requested within 60 days of the creditor's notification.
The rule is written about what the creditor did and what the creditor tells the applicant. It says the reasons must be specific, and that saying the applicant failed to achieve a qualifying score on the creditor's credit scoring system is not enough (1002.9(b)(2)). So a lender that uses an AI model still needs to be able to state the specific reasons. The CFPB withdrew Circular 2022-03 on 12 May 2025.
Sources: Regulation B, 12 CFR 1002.9, on eCFR (https://www.ecfr.gov/current/title-12/chapter-X/part-1002/section-1002.9), and CFPB, Interpretive Rules, Policy Statements, and Advisory Opinions; Withdrawal, Federal Register 2025-08286, 12 May 2025 (https://www.federalregister.gov/documents/2025/05/12/2025-08286/interpretive-rules-policy-statements-and-advisory-opinions-withdrawal), both read 11 November 2026.
security
What should a lender be able to show for an AI credit model?
Ask whether your team could show each of these items to a reviewer. A list of every model that touches a credit decision, with a named owner. The data each model uses and where it comes from. A way to turn a model output into the specific reasons given to an applicant. Records of testing before use and a check after launch. A person who can override the model and a record of when that happened.
If one of these cannot be shown, that gap is the place to start. This list is our suggestion, not a regulatory checklist.
kriv role
Where Kriv helps
Kriv AI helps regulated teams set up the governance around AI systems: the inventory, the ownership, the review steps and the records. For lenders, see our guides to AI model risk management for banks and AI for banks.
Kriv does not build or sell credit scoring models, and does not give legal opinions on lending rules.
limits
What this page could not confirm
This page rests on two sources: Regulation B, 12 CFR 1002.9, and the CFPB withdrawal notice published in the Federal Register on 12 May 2025, both read on 11 November 2026. We did not review state rules, the Fair Credit Reporting Act, federal bank supervisory guidance on model risk, or any vendor scoring product, and we give no figures on accuracy or approval rates. Ask your counsel which rules apply to your lending before you rely on this page.
Related resources
Continue exploring
Straight answers
Frequently asked questions about AI credit scoring: what it is and what a lender has to explain
What is AI credit scoring?
Using machine learning models to help estimate repayment risk and support a credit decision. This is our plain wording, not a legal definition.
Do lenders still have to give reasons for a refusal if an AI model decided?
Yes. Regulation B (12 CFR 1002.9) requires the notice of adverse action to contain a statement of specific reasons for the action taken, or a disclosure of the applicant's right to that statement within 30 days if it is requested within 60 days (eCFR, read 11 November 2026).
Does Kriv build credit scoring models?
No. Kriv AI helps with the governance around AI systems, such as inventory, ownership, review and records.
Is this legal advice?
No. Ask your counsel which rules apply to your lending.
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