AI Regulation and the Case for a Federal Moratorium
TL;DR
Background: Artificial intelligence is rapidly evolving, even as it is integrated into nearly every sector of the U.S. economy. State-level lawmakers have, in response, introduced more than 1,000 AI-related bills in recent years. The question at this point is not whether AI will be regulated, but whether regulation will maximize welfare while preserving America’s competitive advantage.
But… The rush to pass premature and inconsistent state laws threatens to foment massive compliance costs, legal uncertainty, and barriers to entry that stifle innovation and distort markets. This would ultimately guarantee that fewer firms will be able to serve the market with fewer products, paradoxically harming the very competition that regulation is often intended to foster.
However… A core problem with premature regulation is that AI is not a single product, but a vast collection of differing computational methods. When lawmakers try to write rules for such a poorly defined category, they create policies that are either too broad—stifling beneficial uses alongside harmful ones—or too narrow, missing new applications entirely.
A federal moratorium on new state-level AI-specific regulations would offer a period of “regulatory forbearance.” This strategic pause would preempt a costly and inefficient patchwork of rules, allowing time for a coherent and evidence-based national framework to emerge. By relying on existing legal principles and learning from historical precedent, this approach would foster innovation and competition, while ensuring that essential consumer protections remain in place.
KEY TAKEAWAYS
Fragmented Regulation Is Economically Hazardous
The push for state-level AI rules creates significant economic risk, primarily because AI is not a single, easily defined technology. This definitional ambiguity makes it difficult for regulators to craft effective rules, leading to policies that may be either too broad or too narrow. The resulting compliance burdens function as a tax on innovation.
These costs are largely fixed, meaning they disproportionately harm smaller firms and startups by creating significant entry barriers and reducing market competition. Fragmented rules may also impede the network effects and data sharing that are essential to improve AI systems, potentially reducing their quality and effectiveness.
For companies that operate nationally, these compliance costs will tend to multiply. A firm seeking to deploy an AI service across the country cannot simply adhere to one set of regulations; it must comply with the most restrictive requirements across all jurisdictions at the same time. The end result is a regulatory environment comprising the most stringent rules from various states. The economic toll of such regulatory fragmentation is substantial. For example, the lack of a uniform federal privacy law is projected to cost the U.S. economy between $98 and $112 billion annually, or more than a $1 trillion over a decade.
These heavy regulatory burdens function as a direct tax on firms, in turn dampening the incentives to innovate. One study found that such costs effectively act like a 2.5% tax on profits, leading to a 5.4% reduction in aggregate innovation output across the economy. In short, when firms are forced to spend resources to navigate a complex web of regulations, their capacity to invest in new products, services, and technologies is significantly reduced.
Federal Moratorium Provides a Proven Path Forward
Facing the rise of a new commercial-spaceflight industry, Congress passed the Commercial Space Launch Amendments Act (SLAA) of 2004, which implemented a “learning period” that restricted the imposition of new safety regulations on companies like SpaceX and Blue Origin. This strategic forbearance was explicitly designed to prevent regulatory uncertainty from stifling a nascent industry with great potential. By avoiding premature mandates based on an incomplete understanding of the technology and its risks, this approach allowed for tremendous growth and innovation. The current AI landscape presents a striking parallel, suggesting a similar period of strategic patience could yield substantial economic benefits.
Existing Laws Are Already Equipped to Handle AI Harms
The argument that existing laws are insufficient to manage the new risks posed by AI conflates the theoretical possibility of novel harms with the practical necessity for new regulatory tools. To date, the evidence suggests that most purported AI-specific harms are simply new manifestations of familiar legal issues.
A moratorium on state regulation would not leave a regulatory vacuum. A core principle of efficient regulation is “technology neutrality”—the idea that laws should focus on harmful outcomes, not the specific tools used to cause them. The United States already has a robust legal toolkit refined over decades that can address harms caused by AI.
These existing frameworks include tort law, which the principles of negligence and product liability can hold developers accountable for AI systems that cause physical or economic harm.
Various consumer-protection statutes can also remedy misrepresentations about what an AI can do or other unfair and deceptive practices. And while the scale of potential AI-related fraud may be new, the existence of fraud is not. Actions for fraud can be brought whether or not someone uses AI.
Moreover, existing civil-rights laws are designed to address biased outcomes in areas like hiring or housing, regardless whether the discrimination is performed by a human or an algorithm.
The challenge is not a lack of relevant laws, but the need to effectively enforce these existing, technology-neutral frameworks in new contexts. Relying on this existing legal toolkit has significant economic advantages over crafting new AI-specific rules. Established laws have been refined over decades, reducing the risk of unintended consequences, and they apply consistent standards that promote fair competition. This approach primarily uses ex-post liability—providing remedies for actual harms after they occur—rather than ex-ante regulation, which allows market forces to drive innovation.
The pattern is clear: when we examine purported “AI-specific” harms closely, they consistently reveal themselves as traditional legal problems with a technological twist. Algorithmic bias in hiring, for example, is fundamentally a discrimination issue that existing civil rights laws are already designed to address, even if the discriminatory tool is an algorithm instead of a human manager.
Furthermore, when truly novel problems on the scale of existential risk are considered, these are, by their very nature, matters of national and international importance. Such issues are far more appropriate for Congress to consider as part of a holistic view of the economy and national security, rather than being addressed through a disjointed patchwork of state laws.
For more on this issue, see Kristian Stout’s “Federal Preemption and AI Regulation: A Law and Economics Case for Strategic Forbearance.”