Assessing U.S. Financial Regulatory Design: Lessons for AI - Guest post by Jeremy Kress

As part of our series on AI Risk Management and Financial Regulation, University of Michigan professor Jeremy Kress brings concepts from financial regulation to AI.

AI risk management and financial regulation

We publish this post as part of our series joining insights from AI and financial regulation (see here, here, here, and here). This post comes to us from University of Michigan Ross School of Business professor Jeremy Kress.

It is often said that no one would ever design a system like the United States’ financial regulatory apparatus from scratch. With numerous state and federal agencies whose jurisdictions frequently overlap and conflict, the U.S. financial regulatory architecture defies coherent explanation. And in fact, this byzantine system is not a product of intelligent design. Instead, it developed over time through accretion, as policymakers responded piecemeal to financial crises and new developments in financial markets.

This blog post provides an overview of the United States’ bank regulatory system and highlights several key lessons that could help inform the development of regulatory approaches for AI. (The regulation of nonbank financial companies, and its implications for AI, is deserving of a separate blog post.) This post identifies the strengths and weaknesses of the United States’ fragmented approach to regulating banks and offers suggestions for how AI policymakers can leverage the positive aspects of the U.S. bank regulatory system while avoiding the pitfalls.

Consistent with its federalist traditions, the United States has a “dual banking system.” That is, a bank that operates in the United States can be chartered by any one of the 50 states, or it can be chartered by the federal government. This dual system can be beneficial in many ways: state regulators are often more attuned to local economic conditions and may be better equipped to address small banks’ needs, while competition among state and federal regulators potentially leads to more efficient and responsive regulation. At the same time, however, the dual banking systems has evident downsides. The proliferation of multiple regulatory frameworks can create confusion and complexity, while so-called “charter shopping” enables banks to pit authorities against one another to obtain lighter-touch regulation.

Even at the federal level, the U.S. bank regulatory system is remarkably fragmented. Every U.S. bank must have a federal regulator, regardless of its charter type. A bank chartered by the federal government is regulated by the Office of the Comptroller of the Currency (OCC), while a state bank may choose to be regulated by the Federal Reserve System (Fed) or the Federal Deposit Insurance Corporation (FDIC). (In addition, any company that owns a bank—a bank holding company—is regulated by the Fed.) This jurisdictional split can allow for helpful specialization. In general, large banks tend to gravitate to the OCC, while smaller banks usually prefer the Fed and FDIC because of the unique expertise that each of these agencies have developed over time. On the other hand, however, jurisdictional fragmentation at the federal level has notable costs. Maintaining multiple bank regulatory agencies is costly and inefficient, and interagency rulemaking can be difficult, sometimes leading to lax oversight or an unlevel playing field among different types of banks.

The U.S. bank regulatory framework operates against the backdrop of an international standard-setting body known as the Basel Committee on Banking Supervision (BCBS). The BCBS, which consists of central bank officials and bank supervisors, establishes minimum regulatory standards that promote financial stability within and across countries, while maintaining a competitive balance among banks operating globally. This coordinated approach has succeeded in many ways: it facilitates multilateral input on global banking standards while discouraging countries from loosening safeguards below the internationally agreed minimums to attract banking business. Of course, the system is not perfect. BCBS standards are nonbinding, and the BCBS lacks strong enforcement mechanisms. Thus, the BCBS in unable to discipline jurisdictions that do not adhere to minimum standards. Critics also allege that BCBS standard setting is opaque and the use of internationally negotiated standards undermines democratic accountability within the domestic political process.

This quick overview of U.S. bank regulation yields three potential lessons for the nascent regulation of AI. First, since AI technologies operate across state (and, indeed, national) borders, federal regulation is appropriate. The dual banking system worked reasonably well decades ago when banks were limited to operating within a single state. Since banks have expanded across state lines, however, the dual banking system has created challenges. Consider, for example, the U.S. Supreme Court case Marquette National Bank v. First of Omaha Service Corp., which allowed national banks to evade state usury limits by charging credit card interest rates based on the usury laws of the state in which the bank was located, rather than the state where the cardholder lives. While many banks continue to operate only in their local communities and benefit from state regulation, AI’s nationwide reach likely demands a federal overseer to avoid inconsistent state-level approaches.

Second, at the federal level, it would be best to identify (or create) a single agency responsible for regulating AI. Banks sometimes defend the status quo (most banks seem to like their regulators), but many outside observers agree that fragmentation among the OCC, Fed, and FDIC is a net negative for the financial system. Policymakers have long tried to consolidate the U.S. banking agencies, and recent reports suggest that President Donald Trump’s team may renew efforts to rationalize the regulatory system. If past is prologue, however, entrenched interests and regulatory inertia will prevent such reform. AI policymakers could avoid this persistent fragmentation by identifying one federal agency responsible for AI oversight from the outset.

Finally, the U.S. bank regulatory system is a testament to the benefits of strong global coordination. Prior to the establishment of the BCBS, the United States risked losing banking business to countries like Japan that established weak safeguards. By embracing a leading role within the BCBS, the United States has helped standardize the regulation of internationally active banks and, in doing so, helped preserve U.S. banks’ international competitiveness. AI regulators should likewise pursue international coordination through a multilateral standard-setting body. When doing so, policymakers may wish to rectify one of the BCBS’ shortcomings by adopting enforcement mechanisms to discipline jurisdictions that fail to adhere to minimum AI regulatory standards.

In sum, the U.S. bank regulatory framework is far from perfect, but its imperfections can provide important lessons for AI policymakers as they begin grappling with how to oversee new technologies. If the U.S. bank regulatory experience has taught us anything, it is that designing a well-functioning system at the outset is critically important because it is difficult to fix a flawed system over time.