Summary: When a scoring model or any other automated system declines an applicant, Regulation B still requires a written notice that states the specific, principal reasons for the decision. The CFPB’s two circulars on this question were withdrawn in May 2025, so the controlling text is 12 CFR 1002.9 and its Official Interpretations, and that text has not changed. The practical task is to keep, for every decision, the factors the model actually used and the evidence behind them.
ECOA and Regulation B in one paragraph
The Equal Credit Opportunity Act is implemented by Regulation B, 12 CFR Part 1002. Section 1002.9(a)(1) requires a creditor to notify an applicant of action taken within 30 days after receiving a completed application. When the action is adverse, 1002.9(a)(2) says the notice must be in writing and contain the action taken, the creditor’s name and address, the ECOA notice, the name and address of the agency that administers compliance, and either “A statement of specific reasons for the action taken” or a disclosure of the applicant’s right to request one. Section 1002.9(b)(2) sets the standard for that statement: it “must be specific and indicate the principal reason(s) for the adverse action. Statements that the adverse action was based on the creditor’s internal standards or policies or that the applicant, joint applicant, or similar party failed to achieve a qualifying score on the creditor’s credit scoring system are insufficient.”
The Official Interpretations in Supplement I fill in the rest. Comment 9(b)(2)-2: “The specific reasons disclosed under §§ 1002.9(a)(2) and (b)(2) must relate to and accurately describe the factors actually considered or scored by a creditor.” Comment 9(b)(2)-4, on credit scoring systems: “the reasons disclosed must relate only to those factors actually scored in the system. Moreover, no factor that was a principal reason for adverse action may be excluded from disclosure.”
Nothing in that text depends on how the decision was reached.
What the circulars said, and their status
The CFPB issued two circulars on this subject. Circular 2022-03, published June 14, 2022, addressed complex algorithms directly: “The adverse action notice requirements of ECOA and Regulation B, however, apply equally to all credit decisions, regardless of the technology used to make them.”
Circular 2023-03, issued September 19, 2023 and published in the Federal Register on April 17, 2024, asked whether a creditor using a complex model may rely on the checklist of reasons in the CFPB’s sample forms when those reasons do not match the real ones. Its response: “No, creditors may not rely on the checklist of reasons provided in the sample forms (currently codified in Regulation B) to satisfy their obligations under ECOA if those reasons do not specifically and accurately indicate the principal reason(s) for the adverse action.”
The circular’s example is the one to remember. If a model denies an application “due to an applicant’s chosen profession,” then a notice stating “insufficient projected income” or “income insufficient for amount of credit requested” “would likely fail to meet the creditor’s legal obligations.”
Both circulars are now withdrawn. On May 12, 2025, the CFPB published a Federal Register notice withdrawing many guidance documents, with Circular 2022-03 and Circular 2023-03 on the list: “The withdrawals are applicable as of May 12, 2025.” The notice adds: “Such withdrawal is not necessarily final. The Bureau intends to continue reviewing all guidance documents to determine whether they should ultimately be retained. However, the Bureau has determined that the guidance identified in section III should not be enforced or otherwise relied upon by the Bureau while this review is ongoing.” As of October 2026, the CFPB’s Withdrawn Guidance page still lists both circulars as withdrawn on May 12, 2025, and neither has been reissued.
What the withdrawal did not do is change Regulation B. Section 1002.9 on eCFR carries no amendment after March 20, 2023. The sample-form limits the circulars described come from Appendix C to Part 1002, which is still in force: “The sample forms are illustrative and may not be appropriate for all creditors” and “If the reasons listed on the forms are not the factors actually used, a creditor will not satisfy the notice requirement by simply checking the closest identifiable factor listed.”
Why “insufficient income” can fail
A sample-form reason fails when it is not the factor the model used. Three situations produce that.
The first is the circular’s example: the model acted on a variable such as occupation, and the lender reaches for the nearest income box on the form. Appendix C says the closest identifiable factor is not good enough.
The second is a score-based decline with no record of which inputs drove the score. Comment 9(b)(2)-5 describes acceptable methods for selecting reasons from a scoring system, such as identifying the factors for which the applicant’s score “fell furthest below the average score” of applicants at or near the passing cutoff. Without per-factor contributions for each decision, there is nothing to select from.
The third is a combined system. Comment 9(b)(2)-7 says that where a decision has both scoring and judgmental components, “the reasons for the denial must come from the component of the system that the applicant failed.” If the model passed the file and an underwriter declined it on a document review, the notice must give the underwriter’s reasons, not the model’s lowest factors.
A fictional example. An applicant requests a $5,000 installment loan. The model’s output shows the three largest negative contributions came from months since the most recent delinquency, utilization on revolving accounts, and a thin file of two tradelines. Stated income was $4,200 a month and the model did not weight it. A notice that says “income insufficient for amount of credit requested” names a factor the model did not use. A compliant notice names the three factors that were scored, in plain terms, and no others.
Keeping the reasons and evidence for every decision
A lender cannot write an accurate notice without a record. The record has four parts.
The factors scored. For each decision, store the model’s inputs and each input’s contribution to the outcome, ranked.
The component that declined. Record whether the decline came from the model, from a rule that fires before or after it, or from a person. Comment 9(b)(2)-8 says that when an automatic-denial factor such as a prior bankruptcy causes the decline, “the creditor must disclose that specific factor.”
The evidence behind each factor. Tie each reason to the page or data element it came from: the tradeline on the credit report, the pay stub, the bank statement line. When a reason comes from a consumer report, comment 9(b)(2)-9 says the FCRA disclosure that a report was used “does not satisfy the ECOA requirement to disclose specific reasons.”
The notice itself, mapped to the factors. Keep the final wording next to the factor list so a reviewer can check them against each other. If a reason on the notice has no factor behind it, or a principal factor has no reason on the notice, the notice is wrong. Kept for every decision, the record also supports a population test: pull every decline for a period and check each notice against its stored factors.
How the automated version works
MightyBot is an AI agent platform for lending and finance. On the platform, the notice rules above are written in plain English, and agents check every decline against them: the reasons on the notice are compared with the factors recorded for the decision, each reason is tied to the page or data element it came from, and a notice whose reasons do not match the record is routed to a person before it goes out.
See credit risk evaluation and compliance monitoring for how the decision record and the review are built.
