(July 9, 2021) A slow, methodical approach should be taken to determining the best regulatory scheme to address the use of the developing technology of artificial intelligence/machine learning (AI/ML) by financial institutions, the state system has advised NCUA and the CFPB in a comment letter submitted late last week.
In addition, NASCUS wrote in its comment letter to the agencies, that approach must also incorporate “close coordination and collaboration between federal and state regulators in addition to the continued collection of input from other stakeholders.”
In March, the agencies opened a 60-day comment period (which they later extended by 30 days) for Information about how financial institutions use AI in their activities, including fraud prevention, personalization of customer services, credit underwriting, and more.
NASCUS wrote that, in developing guidance and regulation, the agencies would need to provide “insight on important factors such as choosing appropriate data sets, transparency, explainability, accountability and appropriate assessment standards among other things.”
Noting that any guidance and regulation would need to be risk-based and flexible but not stifle innovation, NASCUS said it strongly encouraged the agencies to develop rules and guidelines that would accommodate the needs of all institutional asset sizes. “Guidance or regulation that is too inflexible or that makes AI/ML cost prohibitive may severely impact smaller institutions’ ability to utilize these tools, which may ultimately result in lost opportunities/advantages for impacted consumers,” NASCUS wrote.
The state system also recommended a common lexicon be developed for use of AI to establish a common understanding of just what the technology (and related concepts) constitutes. “An agreed-upon set of foundational definitions would likely go a long way toward avoiding future confusion among stakeholders related to divergent regulatory guidance and rules issued to address this area,” NASCUS advised.