Before a company’s loan request reaches a room full of bankers, a computer may have already sized it up.
Complete the form to unlock this article and enjoy unlimited free access to all PYMNTS content — no additional logins required.
Subscribe to our daily newsletter, PYMNTS Today.
By completing this form, you agree to receive marketing communications from PYMNTS and to the sharing of your information with our sponsor, if applicable, in accordance with our Privacy Policy and Terms and Conditions.
Artificial intelligence tools now score borrowers, sort through financial statements and write first drafts of loan approval memos. In many cases, nobody at the bank formally signed off on that change. That could become a big problem when someone asks how a loan decision was made.
That’s the warning from Outside GC, a firm that provides outside legal counsel to companies, in an article by Partner Kristin Kreuder. AI often slips into the lending process through the back door. A bank might buy a tool to organize a borrower’s financial data. Over time, the tool starts flagging problems with loan terms. Then it helps write the approval memo. Before long, the bank’s risk ratings start moving with whatever the tool says. Nobody updated the bank’s lending rules to account for it.
The drift creates real legal risk. Banks must be able to explain loan decisions to regulators, borrowers and other lenders, the article said. If a loan goes bad, someone will want to know what the AI reviewed, what it missed and where a person stepped in. A memo that reads like a computer summary may not hold up under that kind of scrutiny.
The rules aren’t simple either. Federal banking regulators updated their guidance on how banks manage risk models in April, according to the article. The guidance leaves out AI tools that write text or act on their own. This doesn’t give banks a free pass. Other rules on safety, privacy and outside vendors can still apply. Federal fair lending law also requires lenders to give specific reasons when they turn down a business applicant. A tool that can’t explain itself makes that tough.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
State laws may come into play, too, the article said. A new Colorado law takes effect Jan. 1, 2027, and covers automated decisions about individuals, including lending. It won’t turn a corporate loan into a consumer one, but it could apply when a business owner personally backs the loan.
AI monitoring brings its own headaches, according to the article. A tool that constantly scans borrowers for trouble creates a dated record of what the bank knew and when. If warnings pile up with no response, that record could surface in a lawsuit or bankruptcy fight. Too many alerts can also backfire. Staff may start treating serious warnings as background noise.
The job for bank lawyers usually isn’t to slow down AI adoption, per the article. It’s being able to account for AI use after the fact.
“The algorithm can help with the analysis,” the article said. “The judgment still belongs to the institution.”
Banks should keep an up-to-date map of where AI shows up in the lending process, according to the article. Loan files should show real human review. When a person overrides the tool, the bank should record it and explain why. Rules for handling alerts should be written before the first one arrives.
AI tools that can take actions on their own need tight limits and detailed logs, the article said. For loans shared by groups of banks, the bank managing the loan may use AI to check whether a borrower is meeting its terms. The other lenders may rely on those results without knowing how they were produced. Loan contracts written for human review may need a fresh look.
Finally, banks should report on AI oversight to their boards in plain English, according to the article. The winners won’t be the banks that adopt AI the fastest. They’ll be the ones that can use it, explain it and defend it.
For all PYMNTS AI coverage, subscribe to the daily AI newsletter.
Tech Companies Battle Infestation of AI-Discovered Bugs
Coinbase Customer Data Exposed in Vendor’s Breach
Payments Leaders Say What Changed Their Minds in 2026
Businesses Believe in Payment Guarantees but Few Actually Use Them
Get PYMNTS Today, AI, B2B and more.
Own the conversation. Drive demand.
About the Author
