How AI Is Used in Loan Decisions and What Your Rights Are

You apply for a personal loan on your phone, and 40 seconds later you have an answer. No loan officer read your file. A model did. For many borrowers that speed is welcome. But when the answer is “no”, the obvious question follows: who decided, based on what, and can I challenge it?

AI now plays a role in a large share of consumer lending decisions, from credit cards to auto loans to small business loans. This guide explains where AI shows up in the process, what “alternative data” means, and the legal rights you keep no matter how the decision was made.

Where AI shows up in a loan decision

“AI lending” is not one system. It is a set of models that each handle a step in the process.

Some lenders use machine learning models that weigh hundreds of variables. Others use traditional scorecards with an AI layer for fraud and document checks. Many large banks blend both. For the bigger picture, see our overview of AI in banking and finance.

Traditional scoring vs AI-driven underwriting

Traditional underwriting leans on a credit score, your debt-to-income ratio and a few rules. AI-driven underwriting adds more variables and finds patterns between them.

Neither approach is automatically fairer. A model is only as good as the data and choices behind it.

What alternative data means

“Alternative data” is any information outside the traditional credit report. Common examples:

• Bank account cash flow: deposits, balances, overdrafts and recurring bills, usually shared with your permission through account linking.
• Rent and utility payments, when reported or verified. Our guide on whether paying rent builds credit explains how that works.
• Employment and income verification from payroll data.
• Education and work history, used by some lenders, which has drawn scrutiny because it can act as a stand-in for protected characteristics.

Cash-flow data can genuinely help people with limited credit history. Fannie Mae, for example, lets its mortgage underwriting system consider 12 months of on-time rent seen in bank data. But more data also means more ways for errors and unfair proxies to creep in.

Key takeaway: Whatever data a lender uses, a denial still has to come with real, specific reasons. “The algorithm said no” is not a legally adequate explanation.

Your rights when AI is involved

US lending law is technology-neutral. The rules apply whether a person, a spreadsheet or a neural network made the call.

Specific reasons for a denial (ECOA and Regulation B)

The Equal Credit Opportunity Act, implemented by Regulation B, requires a creditor to notify you of its decision within 30 days of receiving a completed application. If it takes adverse action, such as a denial or worse terms than you asked for, it must either give you a statement of specific reasons or tell you that you can request them within 60 days.

Those reasons must be specific and point to the principal reasons for the decision. Regulation B states that saying you failed the creditor’s internal standards, or didn’t reach a qualifying score on its scoring system, is not sufficient. In 2022 the Consumer Financial Protection Bureau issued a circular stating that this applies even when the model is too complex for the lender to understand. The CFPB withdrew that circular along with dozens of other guidance documents in May 2025, but the underlying requirement for specific reasons comes from the regulation itself.

Your credit report rights (FCRA)

If a decision was based even partly on a credit report, the Fair Credit Reporting Act requires an adverse action notice that includes:

• The name and contact details of the credit bureau that supplied the report
• A statement that the bureau didn’t make the decision
• Your right to a free copy of the report if you ask within 60 days
• Your right to dispute inaccurate or incomplete information
• The credit score used, if one was used

If you find a mistake, follow our steps for disputing a credit report error.

Protection from discrimination

ECOA prohibits discrimination in any credit transaction based on race, color, religion, national origin, sex, marital status or age, because income comes from public assistance, or because you exercised your consumer credit rights. That ban applies to AI models too.

Fair lending rules shifted in 2026. The CFPB published a final rule on April 22, 2026, effective July 21, 2026, stating that ECOA does not provide for disparate impact liability, meaning claims based on a neutral practice that disproportionately harms a protected group without intent. The rule keeps disparate treatment liability, including the use of neutral factors as proxies for protected characteristics. The rule is being challenged in federal court, and some state laws and other federal laws, such as the Fair Housing Act for mortgages, may still reach unintended discrimination. This is not legal advice.

State AI laws are adding new rules

Several states are writing their own rules for automated decisions. Colorado, for example, replaced its 2024 AI Act with a rewritten law, SB 26-189, signed in May 2026 and scheduled to take effect January 1, 2027, covering the use of automated decision-making technology in consequential decisions about consumers. Other states have AI, privacy or insurance rules that touch lending. Our roundup of US state AI laws in 2026 tracks the main ones.

Because these laws vary by state and are still changing, check your state attorney general’s or financial regulator’s website for current rules.

What to do if an AI system denies your loan

1. Read the notice carefully. Note the specific reasons and which credit bureau, if any, supplied your report.
2. Request the reasons if they weren’t included. You have 60 days to ask if the notice only told you of your right to request them.
3. Get your free credit report from the bureau named, and check it for errors, old accounts or identity theft.
4. Dispute mistakes with both the bureau and the company that reported the data.
5. Check the data you shared. If you linked a bank account, ask whether a miscategorized transaction, like a transfer counted as a bill, hurt you.
6. Fix what you can control. Lower card balances, add positive history, or wait until a recent late payment ages.
7. Shop elsewhere. Different lenders use different models, so one denial doesn’t mean every lender will say no. Our guide on how online personal loans work explains prequalification without a hard inquiry.
8. File a complaint with the CFPB or your state regulator if you believe the reasons were vague, wrong or discriminatory.

Where to start: protect yourself before you apply

• Check your reports first at AnnualCreditReport.com so there are no surprises. Our guide to how credit scores are calculated shows what matters most.
• Prequalify with soft-pull offers before submitting full applications.
• Be consistent. Make sure the income, address and employer you enter match your documents. Mismatches trigger fraud flags.
• Share bank data deliberately. Linking accounts can help if your deposits are steady. If your account regularly overdraws, it may not.
• Keep copies of your application, notices and any chat transcripts with the lender.

Conclusion

AI has made lending faster and, for some thin-file borrowers, more accessible. It has also made decisions harder to see inside. The good news is that your core rights still apply: specific reasons for a denial, a free copy of the credit report used, the right to dispute errors and protection from intentional discrimination. Use them. A clear, specific explanation is the starting point for fixing whatever stood in the way.

This is general information, not legal or financial advice.

Sources

• CFPB: Regulation B, section 1002.9 notifications — 30-day timing and specific-reason requirements
• FTC: using consumer reports for credit decisions — FCRA adverse action notice contents
• Venable: CFPB makes significant changes to Regulation B — April 2026 final rule on disparate impact and effective date
• Fannie Mae: on-time rent payments and cash flow in Desktop Underwriter — use of bank data in underwriting