CFJ Blog Post: Only Markets Can Regulate AI
- Jeffrey Depp
- Mar 26
- 2 min read
No government, federal or state, has the knowledge to do it
Artificial intelligence has quickly become one of the most important policy debates in America. As lawmakers, regulators, and advocacy groups propose an expanding array of state and federal rules, the discussion has largely focused on a familiar question: Should AI be regulated by Washington or by the states?
In a new essay, Only Markets Can Regulate AI, CFJ argues that this debate starts from a flawed premise. The real issue is not which level of government should regulate AI, but whether government institutions possess the knowledge, incentives, and institutional capacity necessary to regulate a rapidly evolving technology in the first place.
Drawing on the insights of economists such as Friedrich Hayek, Ludwig von Mises, and Israel Kirzner, the essay explains why AI presents a profound “knowledge problem” for regulators. AI is not a stable industry with known risks and predictable outcomes. It is a dynamic process of experimentation, discovery, and adaptation. Its most important capabilities, applications, and even dangers are often unknown until they emerge through use. Attempts to impose static rules on such a process risk suppressing innovation, locking in errors, and preventing the very discovery needed to understand the technology.
The essay also examines the institutional realities of government regulation. Public-choice economics teaches that regulators and lawmakers operate under imperfect information and are influenced by political incentives, special-interest pressures, and bureaucratic constraints. As a result, regulatory systems often fail to deliver the safety, fairness, and accountability they promise, while simultaneously imposing substantial costs on innovation and competition.
Rather than viewing AI governance as something that must be imposed from above, the essay highlights the role of market discipline. AI companies already face powerful incentives from users, enterprise customers, reputational pressures, and competitive forces. Firms that fail to meet customer expectations lose market share, while those that successfully address concerns about reliability, safety, and trust are rewarded. These decentralized feedback mechanisms adapt far more quickly than regulatory processes and are better equipped to respond to a technology that continues to evolve at extraordinary speed.
At a time when policymakers are increasingly tempted to “do something” about AI, Only Markets Can Regulate AI offers a different perspective: the most effective system for governing emerging technologies is not centralized control, but the spontaneous order generated by millions of users making choices in the marketplace.
Read the full essay here:
