A Constitutional Yellow Light on a FINRA for AI

Kevin T. Frazier

Many innovators find the growing uncertainty, frequent changes, and secret frameworks in the AI regulatory space as barriers to investment in the research, development, and deployment of new technology. At the same time, there’s broad recognition that laws such as the EU AI Act risk being out of date by their effective date, representing a different constraint on the AI ecosystem. The frustrating reality is that there’s no “right” way to govern AI. Existing laws, including consumer protection statutes, already address many of the most common harms associated with AI. Yet frontier issues and edge cases will require new paradigms and regulatory certainty to allow innovators to fully pursue possibilities while resolving the public’s and policymakers’ concerns about legitimate risks, such as cybersecurity threats.

New governance efforts should be tailored to address cases where clear market failures underpinning broader AI adoption and ongoing AI research occur or where the lack of regulatory certainty prevents important development of the technology. Three examples of such failures include: Inadequate information about model capabilities and values is hindering the ability of consumers to make knowledgeable decisions; insufficient supply of independent AI auditors threatens to prevent the timely and thorough investigation of company practices; and ineffective initiatives to increase the talent pipeline of AI experts may hinder the nation’s long-term tech ambitions.

These and other shortcomings have not been lost on the public and, consequently, policymakers and other AI stakeholders. The mounting pressure for something to be done is generating manifold policy ideas, many of which would result in problematic government actions that could limit the benefits of AI and intervene in competitive and emerging markets. 

Many stakeholders are therefore looking at potential alternatives that could provide balance through industry self-regulation or similar methods that continue to allow competition but could resolve significant concerns and establish industry norms. One such idea is the creation of an organization for AI that would function similarly to the Financial Industry Regulatory Authority (FINRA). In general, a FINRA for AI would mean relying on a private body to oversee the development and enforcement of standards while designating a federal regulator to serve as a backstop only when truly necessary.

However, as I initially explored on the Cato at Liberty blog, FINRA is far from a perfect regulatory body in its own domain. A review of its history exposes that the intended benefits of this regime have often not been realized. Part of those shortcomings arise from the fact that the Constitution limits the extent to which private actors can exercise legislative powers assigned to Congress.

I explore those constitutional redlines in more detail via a new piece in the Yale Journal on Regulation. You can read the full piece here.