MBA: AI Can Originate Mortgages, But Federal Rules Still Require A Human Loan Officer
By CU Today Staff —
WASHINGTON—Mortgage lenders are rapidly embracing artificial intelligence across the loan lifecycle, but federal law still effectively requires a human mortgage loan originator to remain involved in the origination process, according to a new Mortgage Bankers Association white paper prepared by law firm Orrick, Herrington & Sutcliffe.
The report, "Examining AI-Powered Mortgage Through the Lens of Federal Law," found that lenders are increasingly deploying generative, predictive and agentic AI tools for customer engagement, underwriting, fraud detection, servicing and compliance functions. Some industry participants believe advanced AI systems may soon be capable of conducting end-to-end mortgage originations, from application intake to underwriting and document preparation, with little or no human intervention.
Despite those advances, the paper concludes that AI systems themselves do not need mortgage loan originator licenses because the SAFE Act applies to human "individuals," not software models. At the same time, federal Truth in Lending Act and Regulation Z disclosure requirements effectively require a human loan originator with a Nationwide Multistate Licensing System identifier to be assigned to each mortgage transaction and disclosed to borrowers.
The white paper warns lenders against removing humans entirely from the process. It argues that naming a loan officer as a borrower's primary contact while conducting the entire origination through AI could create potential unfair or deceptive acts and practices risk if consumers reasonably expect that person to be involved in the transaction. The authors recommend maintaining a meaningful "human in the loop" role and ensuring borrowers can reach a licensed loan officer when needed.
MBA also said the industry should move quickly to establish a common AI governance framework before regulators and states impose a patchwork of conflicting requirements. The report points to recent actions by Freddie Mac and Fannie Mae requiring seller-servicers to establish AI governance and risk-management policies, while noting that federal lawmakers have provided limited guidance on how AI should be used in mortgage lending.
Among the biggest risks identified are fair lending compliance, explainability of AI-driven credit decisions, potential steering concerns, data privacy, cybersecurity and vendor oversight. The paper recommends robust testing for bias, frequent model validation, clear governance structures and transparent disclosures when consumers interact with AI systems, particularly in servicing and loss-mitigation functions.
Originally reported by CU Today.