MISMO Summit: A Look at the AI Legal Landscape
(l. to r.: Gabriel Acosta, Liz Facemire)
WASHINGTON, D.C.–Innovations around artificial intelligence (AI) such as Generative AI are exciting, but there are important federal, state and local legal legislative and regulatory issues to consider, policy analysts said here at the MISMO Fall Summit.
“Let’s start by talking about what we’re talking about when we say AI,” said Gabriel Acosta, Regulatory Specialist, Residential Policy & Strategic Industry Engagement with the Mortgage Bankers Association. He said there are numerous definitions at both the state and the federal level, but lawmakers have mostly settled on this definition: AI is a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations or decisions influencing real or virtual environments.
“As you may realize, this definition is incredibly broad and includes things like spell checking. It’s so broad that Colorado specifically defined AI to exclude certain mundane tools like spell checking, calculators and spreadsheets,” Acosta said. “Given this wide definition, although you may not know it, AI is actually used a lot in the mortgage space.”
Mortgage Bankers Association Director of State Government Affairs Liz Facemire noted this definition would include automated underwriting systems, automated valuation models and credit scoring models. “Really, a ton of information and data that we use goes into what [the states] would define as artificial intelligence,” she said.
Acosta and Facemire discussed the legislative and regulatory activity they are seeing at the federal level and in various states.
“On the federal level, there’s been a lot of smoke, but no fire. The federal government is only really just beginning to grapple with how to regulate AI, and has really only published reports or non-binding risk management frameworks, so they are a little bit behind the states,” Acosta said.
Acosta noted the White House AI Bill of Rights shows the White House is concerned about protections against discrimination, protections for data privacy and some sort of opt-out manual process. The White House has also issued an executive order instructing federal agencies to start addressing AI issues.
“We’ve seen them so far create a final rule on reconsideration of value, and the Consumer Financial Protection Bureau released an advisory opinion on comparison-shopping tools that start to address the AI issue,” Acosta said. “However, it’s important to note that both of those came before the [White House] executive order, so we’re still waiting to see the fruits of that.”
In Congress, the House Financial Services Committee released a report from its AI working group that identified similar concerns to the White House’s AI Bill of Rights. It focused on the need for anti-discrimination and data privacy protections, Acosta said.
“I would say the CFPB is the furthest along on AI,” Acosta noted. “Two years ago, they released a circular saying that creditors cannot just rely on the output of a complex algorithm as a reason to deny applicants. They need to list the specific reason, so this is kind of the first step–and it’s a pretty significant step–toward addressing AI.”
But most of what’s going on is at the state level, Acosta and Facemire agreed.
Facemire showed a map indicating that more than half of the 50 states have taken up a form of AI legislation on their own. “They see Congress really stalling on almost everything, so they think, I’ve got the right to do it myself. I’m going to go ahead and tackle this AI issue,” she said. “So it’s a very, very active map for 2024, and we’re expecting a lot more in 2025.”
