The introduction of AI policies varies across the mortgage industry, often dictated by the institution type or size, but the rate of adoption means the structured rules will eventually become more of a need and not an option, leaders say. 

Currently, some of the largest nonbanks and other financial institutions have created their own rules to protect their own interests and keep all parties in line. Banks and other depository institutions are frequently governed by strict regulations, leading them to enact technology policies governing AI usage as well.

On the other side, though, small mortgage shops with more limited tech knowledge have few rules in place, sometimes leaving oversight to the best judgment of team members or choosing to limit adoption.

Regardless of where companies sit with their plans today, an AI policy is a smart move that can build trust and serve to advance the company's overall goals, according to Grace Brasington, executive managing director of compliance consultancy firm Asurity Advisors. The firm serves financial and lending institutions of various sizes up to large and super regional banks. 

"There needs to be a baseline understanding from a risk framework standpoint.  How are we being impacted internally and externally?" she said.  

While no policy could cover every scenario, "it needs to incorporate your guiding principles. It needs to have what we consider ethical guidelines for using it and what's acceptable," she added.

A policy for all parties

To be effective, though, AI policy needs to apply not only to employees, but every individual that touches a company's tools and platforms, including third-party vendors and broker partners. 

Knowingly or not, some lenders today approach artificial intelligence with a build-the-plane-while-flying approach, not always aware of potential repercussions for noncompliance, said Jason Bressler, chief technology officer at United Wholesale Mortgage. 

"They're expecting that the vendors they have will protect them without asking any questions around it," he said. 

As a public technology-focused corporation with its own proprietary platform in use, UWM has to take a more intentional approach. "We have AI steering committees. We have AI ethics guidelines. So we have all of that because we're build versus buy," Bressler added.

Still, no matter their size or level of AI savviness they already possess, the same underlying principles will apply to all enterprises. For the tech wary, ignorance is unlikely to pass muster as an excuse if a snafu occurs.

"Based on the regulatory environment that we all live in, you cannot outsource your risk management responsibility," Brasington said.    

"We have dealt with institutions that have relationships with the fintech community, and again, that comes back to making sure that if they're aligned with those fintech partners, there is an understanding of how the tools work from an overall model perspective, because otherwise it would create risk for both parties."

Among some companies in the mortgage industry, though, hesitancy toward using any new technology, much less coming up with forward-looking policies that govern it, is seemingly ingrained. Behind the tech-averse hesitation is often the worry that the tools or their vendors will run afoul of regulators, according to Jodi Hall, president and CEO of The Mortgage Collaborative, an industry network of lenders and service providers.

"Some lenders are so worried about the compliance and about the risk that they do nothing," she said.

Instead, when faced with new digital tools, the mortgage industry has long had a tendency to wait and see how the largest companies and partners act, later basing their own decisions off the outcomes. 

"I think we're in a wait-and-see environment to see what the requirements are going to be from the larger investors, larger mortgage lenders and their adoption of state and federal policies," Hall said. 

That type of caution might not serve lenders well in the long run, though, leading them to play catchup or lose the AI race altogether. They will also see higher costs to originate, "almost pricing themselves out of the market from the unwillingness to try to weed through all of the complexity of navigating policies and compliance," Hall added. 

"If I'm an independent mortgage bank and I'm selling loans to a larger investor, I'm going to start asking the questions of those larger investors about what their policy is and try to get several," Hall suggested. At that point, a lender can weed through all of them to see what might be the strictest and come up with one of their own. 

The importance of human oversight

As it is with many regulations governing data and technology, the value of developing clear AI guidelines stems from consumer-privacy concerns, which are always top of mind among lenders seeking compliance. 

While artificial intelligence is quickly making huge strides in the mortgage industry since arriving just a few years ago, the decisions in the process, from what information gets fed to AI tools to loan approvals still lies with the humans behind it. 

"The ability to synthesize a vast amount of information and to bring it together in a way that's meaningful — it's a beautiful thing, but you can never forget about that human element," Brasington said

Popular platforms like ChatGPT or Anthropic's Claude carry with them risks that a user might share nonpublic personal information on public versions of the platforms accessible by millions. When first testing generative AI platforms, the consequences of sharing the data might not be obvious if clear guidelines aren't in place. 

"You don't want to do anything that would negatively impact the privacy of the consumer by inadvertently sharing it because it's a free tool and it's available to you," Brasington said.

"It doesn't matter how big or small you are. From a reputation and risk perspective, you don't ever want to be in a situation where you allowed NPI data to leave the walls of the institution."

Companies can head off any trouble by keeping humans in the process who raise the warning flags and put up the guardrails and create processes to keep them in check, Hall said.

"I don't think that people should be scared, and I think that they can start small in AI and still be compliant," she said.

If origination volumes increase as expected over the next two years, the mortgage industry should expect to see AI adoption grow as lenders try to deal with the demand. Then, the need to create firm institutional rules will become apparent, Bressler predicts. 

"What we're going to see is a lot of these smaller companies just turn to anybody, and that's when we're going to, I think, start to see policy come into place. Because then, it's going to be a very widespread issue," he said.

"We'll start to see these companies really needing to figure out some way to adopt AI. They're just not going to really be sure exactly how."