The Mortgage Bankers Association (MBA) is urging the mortgage industry to develop a unified framework for managing artificial intelligence as lenders increasingly deploy AI tools across origination, servicing and customer engagement.

In a white paper released Wednesday, the association said AI technologies are rapidly becoming embedded throughout the mortgage process, from customer service chatbots and fraud detection systems to underwriting and servicing operations.

The paper argues that while AI offers significant efficiency gains, the industry faces growing uncertainty about regulatory expectations and legal compliance.

The paper, prepared for the MBA by law firm Orrick, Herrington & Sutcliffe, examines how existing federal laws apply to AI-powered mortgage lending and outlines best practices for lenders that adopt the technology. It also provides an up-to-date overview of MBA members’ engagement and implementation around AI use and regulation, while also posing and analyzing key legal questions about the use of AI.

“AI’s assistance with — and, in some cases, performance of — a broader range of mortgage-related tasks raises novel questions about expectations for human involvement with AI models, as well as risk management more broadly,” the report explained.

SAFE Act ambiguity

MBA noted that mortgage companies are increasingly exploring generative AI, predictive AI and agentic AI systems, with many lenders already using AI-powered chatbots to answer simple questions or support servicing functions. But the paper also said that more advanced systems may soon be capable of handling nearly every stage of the mortgage process.

The association noted that while existing laws like the Secure and Fair Enforcement for Mortgage Licensing Act of 2008 (“SAFE Act”) exist to establish a nationwide licensing system and ensure loan originators are subject to regulatory oversight, the act does not answer whether mortgage companies can offer completely human-free loan originations.

“While it does require human MLOs to be licensed or registered (depending on the nature of their employment), the SAFE Act does not require AI tools to have their own MLO license or registration to engage in loan origination activities,” the report reads.

MBA argued that current federal disclosure requirements effectively require lenders to assign a human mortgage originator to every transaction. Under the Truth in Lending Act and Regulation Z, lenders must disclose the name and Nationwide Multistate Licensing System (NMLS) identification number of an individual loan officer associated with the loan.

The report recommends that lenders maintain a “human in the loop” approach, ensuring a licensed mortgage originator remains available to borrowers and participates in some level of oversight, even when AI performs substantial portions of the origination process.

MBA also warned that lenders could face risks under federal and state consumer protection laws if borrowers are led to believe a human loan officer is overseeing their application when the process is handled entirely by AI.

GSE guidance already in place

The white paper comes as regulators and investors have begun to address AI governance. Earlier this year, Freddie Mac updated its seller-servicer guide to include AI and machine learning governance requirements, while Fannie Mae issued guidance calling on lenders to establish policies and procedures that govern AI systems.

MBA said federal policymakers have provided limited guidance on how existing lending and consumer protection laws apply to AI-enabled mortgage processes. To address that uncertainty, the association is advocating for a principles-based AI risk management framework tailored to the mortgage industry. This would include standards for governance, model validation, fair lending tests, explainability, data privacy and vendor oversight.

MBA also encouraged lenders to implement robust testing programs to monitor AI systems for disparate treatment or disparate impact on protected groups and to maintain documentation showing compliance with fair lending requirements. The report identified several risks associated with AI adoption, including potential fair lending violations, bias in automated decision-making, improper steering of borrowers to certain loan products and consumer privacy concerns.

The association said lenders should prepare for evolving regulatory expectations while also engaging with lawmakers and regulators as states consider legislation governing AI use in financial services.

“The rapid adoption of AI across the mortgage industry presents both significant opportunities and complex legal and regulatory challenges,” the report concluded. “The absence of comprehensive federal and state guidance on AI in mortgage lending creates an imperative for the industry to develop and adopt a unified, principles-based risk management framework.”

This article was written by Sarah Wolak and generated with the assistance of HousingWire Automation, then reviewed by a HousingWire editor before publication.

More: