The Federalist’s Dilemma: State AI Regulation & Pathways Forward
- Author(s):
- Evangelos Razis, James C. Cooper
- Posted:
- 06-2025
- Law & Economics #:
- 25-07
- Availability:
- Full text (most recent) on SSRN
ABSTRACT:
AI has captured everybody’s imagination, especially policymakers. The extent to which imagination has translated into action, however, is a mixed bag. At the federal level, Congress has studied the issue, weighed grand proposals, and held countless hearings on AI but has enacted only modest legislation. While executive branch agencies and the FTC have talked a big game, their accomplishments have also been modest, mostly due to limits on legal authority. Not surprisingly, as with data privacy, states have stepped into the vacuum created by federal inaction with AI regulations of their own. Typically, states acting as laboratories is a good thing, allowing experimentation and competition to hone the efficiency and fit of regulatory regimes to different situations. But when the subject of regulation is interstate – and in this case global—by nature, a patchwork of state regimes is far from ideal. The solution to this dilemma is often seen as a binary: allow the state patchwork to evolve for better or worse, or stop it in its tracks with a federal preemptive response. We see this as a false choice and offer two potentially better paths. First, would be for Congress to enact a national “moratorium” on state laws regulating AI. We argue that this as a superior approach because it will arrest potentially harmful regulation and the patchwork problem and alleviate pressure on Congress to pass premature AI laws merely to prevent the states from acting. Second, would be to honor choice of law provisions in AI-related contracts, thereby fostering competition among firms and states to provide efficient AI regulation. Borrowing from the ideas of Larry Ribstein and various coauthors, we argue that firms would compete for consumers by choosing to be regulated by the regime that maximized their profits, and states would compete to enact efficient laws. In sum, we think the current rush to regulate AI, whether at the state or federal level, is premature. Regulators have existing tools to address consumer harms. The problem is that our federal system, just like nature, abhors a vacuum, and states are filling it with a patchwork of potentially onerous and inconsistent AI requirements. The pressure to prevent state action, in turn, may force Congress’ hands into an ill-considered and hasty response that is little better than the states’ alternative. We see our hybrid approaches as a way out of this dilemma.