CFPB guidance on chatbots, in addition to the directive on denials, have made vendors cautious, and "that's why I feel that we're kind of at a point where we've got to watch how we use AI as an industry," he said while speaking on a conference panel about its role in underwriting.

But while this may make the industry move a little more deliberately when it comes to development and use of the technology, it hasn't stopped progress altogether.

"There are a ton of applications we're working on that include helping the people that have to understand the seller guides," said Poeisz, referring to rules government-related loan buyers set for lenders. "That knowledge to users is certainly a good application of generative AI."

Other underwriting-related technologies that are moving forward with some regulatory scrutiny include digital bank and rent data that can serve as an alternative way to qualify borrowers who lack traditional credit histories.

Oversight agencies are very protective of the use of this consumer-permissioned data. Stakeholders participating in the Federal Housing Finance Agency's TechSprint discussions in July told the agency they see a utility model as one potential long-term outcome.

There is room to move within the rules in this area, Lucky Sandhu, president and CEO of Reliance Financial, told conference attendees while speaking on a panel about alternative credit's potential to grow loan pipelines.

"Regulators will work with you as long as you understand the foundation and fundamentals very, very strongly, especially when it comes to understanding credit defaults and credit risk," Sandhu said.

Alternative credit's potential reach is sizable, said David Battany, executive vice president, capital markets, Guild Mortgage, citing Consumer Financial Protection Bureau data indicating over 50 million adults have insufficient or no traditional credit history.

While alternative credit has long existed, it's been unwieldy to use, with few people willing to go through a process, he noted. But digital advances in consumer-permissioned bank and rent data at government-sponsored enterprises Fannie Mae and Freddie Mac are improving access.

"The GSEs have really taken the lead on this. Also the private market —  the non QM market  — has really innovated in a lot of areas," Battany said. Digital tax-transcript data in particular has been used to qualify self-employed borrowers for the latter product.

While conforming lenders are able underwrite self-employed borrowers, the loans have restrictions. That ends up pushing many into non-qualified mortgage products where lenders have less assurance of compliance with the Consumer Financial Protection Bureau's ability-to-repay rules.

While the enterprises have offered lenders limited relief from representation and warranty risk when digital data validates information on loans submitted for sale in some cases, Fannie has warned whether the information is ATR compliant is a separate question.

And the number of alternative credit borrowers making it through into the GSE market has been limited, according to both Battany and another panelist, Patrick Tadie, executive vice president, global capital markets, structured finance, at Wilmington Trust

One hurdle to the use of alternative credit data by the private credit market is that the rating agencies that have a hand in secondary market pricing consider it to be limited given the small amount of loans originated and their performance track record.

"We still need more data," said Tadie, noting that the view the rating agencies have of it makes originating loans for sale into this market relatively more costly.

Wilmington's parent company, TD Bank, does have a private loan product based on alternative credit that it holds in portfolio rather than selling to the secondary market. But its reach is limited, Tadie said, noting that underwriting requires a lot of compensating factors.

"It's incredibly conservative," he said.