Why AI Won’t Eliminate All Market Failures

AI will slash transaction costs and make markets hum more efficiently—but it will also make some of the economy’s worst problems cheaper to create. From fraud that scales at near-zero marginal cost to infrastructure buckling under AI-driven demand, the technology is a double-edged sword for market failures. This paradox challenges the prevailing optimism that AI will automatically cure inefficiencies, forcing regulators and companies to confront new forms of externalities, information asymmetries, and public-goods challenges.

James Broughel, an economist at the Mercatus Center, argues in a new Forbes analysis that while AI lowers search and coordination costs—classic transaction costs that impede perfect markets—it simultaneously reduces the cost of producing harmful outcomes. Fraudsters can deploy generative AI to craft convincing phishing emails at scale, while automated trading algorithms can amplify flash crashes faster than any human could react. The result: some market failures get smaller, but others get bigger, and entirely new ones emerge.

Broughel’s piece, published June 8, 2026, taps into a growing debate among economists and technologists. For decades, the standard view has been that digital technology, including AI, would push markets closer to the frictionless ideal of perfect competition. Lower transaction costs should, in theory, eliminate many failures rooted in information asymmetry and high search costs. But AI’s ability to produce outputs—both beneficial and harmful—at near-zero marginal cost flips the script. Negative externalities become cheaper to generate, not more expensive. AI market failures, Broughel warns, are not just the old problems in new clothes; they are qualitatively different.

Key evidence lies in recent incidents. In 2025, a series of AI-generated deepfake voice scams cost US businesses over $500 million, according to the FBI. Meanwhile, data centers powering AI models are projected to consume 8% of US electricity by 2030, up from 2% in 2024, straining power grids and raising carbon emissions—a textbook negative externality. On the public-goods front, open-source AI models can be freely copied and misused, making it harder to internalize safety costs. Broughel cites the tragedy of the commons: no single actor has an incentive to limit AI’s harmful spillovers.

Analysis: The broader implication is that AI requires a rethinking of market-failure theory. Traditional remedies—taxes, subsidies, regulation—may need to be supplemented with new tools like algorithmic auditing, liability rules for AI-generated harms, and real-time monitoring of infrastructure stress. As economist Tyler Cowen has noted, AI could create a world of “cheap fraud and expensive trust,” where the cost of verification rises even as production costs fall. Policymakers in the US, EU, and China are already grappling with these issues, but Broughel’s piece suggests they are only scratching the surface.

Outlook: Expect a wave of research and regulation targeting AI-specific market failures. The EU’s AI Act begins enforcement in phases from 2025, but it focuses on safety and bias, not economic externalities. Future milestones include the US Federal Trade Commission’s upcoming report on AI and market competition, due late 2026, and likely court cases over liability for AI-generated fraud. The conversation is shifting from “Can AI fix markets?” to “How do we fix AI’s impact on markets?”—and that question will define economic policy for the next decade.