The prospects for a bill that would defer regulatory enforcement in the interest of encouraging the use of artificial intelligence in the financial services sector were elevated following the House AI task force’s landmark report and selection of a new chair for the House Financial Services Committee.
Among the recommendations of the bipartisan AI task force report released Dec. 17, lawmakers said Congress should “consider the merits of regulatory ‘sandboxes’ that could allow regulators to experiment with AI applications.”
The recommendation comes as consumer advocates, including those at the Consumer Financial Protection Bureau, the Consumer Federation of America and the National Fair Housing Alliance are wary of efforts to suspend the enforcement of existing laws in the sector toward incentivizing innovative use of AI, and as states like Utah are implementing their own legislation.
CFPB particularly warned about the industry’s reliance on opaque “black-box” systems in the context of laws requiring financial lenders to provide specific reasons for their decisions if there is a rejection, or “adverse action,” for example, in Aug. 13 comments to the Treasury Department.
In a statement hailing the task force report, Rep. French Hill (R-AR), highlighted his sponsorship -- along with Rep. Ritchie Torres (D-NY) and Sens. Mike Rounds (R-SD) and Martin Heinrich (D-NM) -- of the Unleashing AI Innovation Act.
Hill, who won a Dec. 12 vote to lead the Financial Services Committee in the new Congress, said the legislation would “[establish] regulatory sandboxes for AI experimentation at various financial government agencies.”
“Private-sector firms should disclose and document how AI is being utilized when using AI in a customer facing or back-office efficiency activity,” Hill also noted. That dovetails with the report’s call for transparency, particularly in sectors like finance and healthcare.
“Company management should be able to ‘show their work’ and certify to their internal and external compliance officials that they are in full compliance with U.S. laws and regulations,” Hill added.
Adam Rust, the director of financial services at CFA, and Michael Akinwumi, chief AI officer for the NFHA and a Rita Allen Civic Science fellow, have told Inside AI Policy that’s not their reading of what Hill’s bill, specifically, would do, with Akinwumi saying the legislation would inappropriately and unnecessarily treat consumers “like guinea pigs.”
But the consumer advocates, along with industry institutions, have also been appealing to the CFPB to issue clearer guidance on what might constitute “Less Discriminatory Alternatives” when considering the use of algorithms to determine who should qualify for credit, or a house loan, for example.
And while Akinwumi is not a fan of the leading federal proposal, he’s pointed to a “sandbox” bill in Utah as presenting a more reasonable approach to encourage innovation.
Both the Utah bill and the Hill bill would solicit applications from industry entities to be participants in a regulatory lab, where they can negotiate alternate terms or plans for complying with finance sector regulations as they try to incorporate AI into their operations.
The Utah bill includes extensive criteria for applicants to qualify, sets strict limits on how long their “regulatory mitigations” can be used, and notes, “A participant remains subject to all legal and regulatory requirements not expressly waived or modified by the terms of the regulatory mitigation agreement.”
By contrast, among much else, the federal bill would make it harder to reject industry entities looking to participate in the regulatory sandbox, and block enforcement of all existing regulation except as laid out in the alternative compliance plans of the applicants.
It would also prevent regulators from taking enforcement actions against an entity within 30 days of an application being filed and allow applicants to continually submit applications as long as they are not “substantially similar.”
“If the financial regulatory agency needs additional time, the agency may vote to extend the application deadline by 90 days,” the federal bill reads. “After the expiration of the 90-day period, if the agency has not made a determination on the application, the application will automatically be deemed approved and effective.”
Akinwumi said the value of the Utah bill would be realized in, “thinking about affirmative defense, there’s an incentive for you to actually, proactively remedy any violation that you see during that [limited trial] period, so you can reduce the fines for violations, and have cure periods before the fines are imposed.” He emphasized entities “would still be fined,” that they “would still be penalized,” adding “that's a model that the sponsors of this [Hill] bill can consider.”
“But total suspension of the law,” he said, “that's something that should never happen.”
Industry representatives from the financial sector weighed in with support for both Hill’s ascension and the task force report.