How Does Agentic AI in Mortgage Origination Meet Safe Act Standards? Hint: It Doesn’t

The SAFE Act (Secure and Fair Enforcement for Mortgage Licensing Act of 2008) is a federal law that requires all residential mortgage loan originators (MLOs) to be licensed or registered through the Nationwide Multistate Licensing System (NMLS). 

It mandates “background checks, credit reports, fingerprinting, pre-licensing education, testing, and annual continuing education.” The goal is to “enhance consumer protection, improve tracking of MLOs, reduce fraud, and ensure minimum standards across states.” 

Regulators interpret the SAFE Act as prohibiting any contact between a mortgage lender and a consumer that involves taking a residential mortgage loan application or negotiating the terms of a residential mortgage loan. 

Communications or interactions cross into "engaging in the business of a residential mortgage loan originator" when they involve more than purely administrative or clerical tasks such as collecting names, area of interest (purchase/refinance), addresses, and contact information. Examples of activities that require an MLO license/registration include: 

1. Soliciting or communicating an offer to a consumer to apply for or obtain a residential mortgage loan (e.g., discussing specific loan programs, rates, or eligibility in a way that promotes origination). 

2. Negotiating or discussing loan terms with a consumer (e.g., interest rates, fees, repayment options, or modifications to proposed terms). 

3. Taking or assisting with a loan application from a consumer (e.g., helping complete forms, collecting personal/financial information beyond basic clerical intake, or advising on what information is needed for approval). 

4. Representing to a consumer (through advertising, phone calls, emails, social media, in-person discussions, or other means) that the individual or their company can originate, arrange, or negotiate a residential mortgage loan. 

5. Communicating on behalf of an MLO that a written offer or counteroffer has been made regarding loan terms (in some interpretations under state rules aligned with SAFE). 

Any initial or substantive written/oral communication that goes beyond gathering basic consumer and transaction data and involves offering, advising, or discussing terms of the loan itself must be through a licensed individual. These apply across channels, including phone calls, emails, texts, social media posts or direct messages, websites, advertisements, in-person meetings, or mail.  

Now that Agentic AI has hit the mortgage industry with consumer facing bots not only taking information but responding to questions and even in some cases conducting an automatic qualification assessment, the question has arisen, is this unlicensed activity in violation of licensing regulations?  Seems clear that it is because a bot cannot be a licensed individual under the act. 

Why not?  Because a technology platform is only as good as it is designed, programmed, and monitored.  It is not a living, organic thing.  In addition, a bot is subject to programming errors and machine learning errors that could result in false, inaccurate or even damaging consequences with respect to the collection, interpretation, management, and evaluation of the data consumers provide it.  Sounds like the perfect environment for violations of Unfair, Deceptive, or Abusive Acts or Practices (UDAAP). More importantly, at least for now, there is no exemption for a tech bot from the SAFE Act for licensing and NMLS registration requirements. 

What about disclaimers and off ramps? When using Agentic AI, should consumers have to approve? Should they be given the right to opt out and only communicate with a human?  Will consumer reluctance to embrace AI, result in lost business opportunities?  These are operational issues and risks that certainly must be considered. 

Some lenders are already using these bots, so what is the big deal? The concern and the risk involve the novelty of the technology.  Auditors, examiners, and enforcement agents are still getting their hands and heads wrapped around the issue.  This is not tacit consent or approval of the use of these platforms. As soon as a complaint arises where a consumer is given false, deceptive, inaccurate, misleading or hallucinogenic responses which result in a blown deal, blown rate lock, or inconsistent rates, terms or fees at closing, the dam will break. 

Is the technology interesting? Yes.  Is it a time saver and cost saver? On its face, yes. Is the use of the technology risk free? The answer is no. For the time being, consider how AI could conceivably improve internal operations and efficiencies.  With respect to Agentic AI the risk of penalties, fines and lawsuits may be too high in comparison to the ROI.