Regulatory Markets
A competitive market of licensed private regulators who sell regulatory services to AI companies, with governments setting required outcomes while market forces drive innovation in compliance.
What it is:
Regulatory markets would require governments to define regulatory outcomes while licensed private regulators compete to develop and sell compliance services to the companies being regulated. Rather than writing detailed technical rules that rapidly become obsolete, governments would specify the results they want — safety thresholds, transparency standards, accountability mechanisms — and license private firms to certify that companies meet those standards. The private regulators compete on price, quality, and the efficiency of their compliance methods, while facing license suspension or revocation if the companies they certify cause harms or fail to achieve required outcomes. This addresses two deficits in current AI governance: the technical deficit (governments lack the expertise to write and update detailed AI regulations) and the democratic deficit (industry self-regulation lacks public accountability and enforcement).
Applied to AI's economic impacts, regulatory markets could operationalize obligations that are easy to state but hard to enforce through conventional regulation. A government might require that firms deploying AI provide meaningful transition support to displaced workers, maintain human oversight of consequential decisions, or ensure that productivity gains are shared rather than captured entirely by shareholders. Translating these principles into auditable, enforceable standards is technically demanding; it requires understanding specific AI systems, workplace contexts, and labor market dynamics. Private regulators competing for business would have incentives to develop sophisticated compliance methods, while the threat of license revocation would align their interests with public outcomes rather than with the firms paying for their services.
The challenge:
The model's viability depends on whether governments can effectively oversee the overseers. Private certification regimes have failed catastrophically when regulators became dependent on the firms they certified, faced inadequate government scrutiny, or were shielded from liability for failures. If private AI regulators compete primarily on cost rather than rigor, the market dynamic could produce a race to the bottom. There is also a question of whether outcome-based regulation is feasible for diffuse economic harms: it is easier to verify that a bridge did not collapse than to verify that automation gains were "meaningfully shared" with affected workers.
Recommended Reading:
Real-world precedents:
Credit rating agencies functioned as private regulatory intermediaries in financial markets but failed catastrophically in 2008 due to inadequate government oversight and deliberate shielding from liability.
Similarly, the FAA's delegation of safety certification to Boeing, which contributed to the 737 MAX crashes, demonstrates the dangers of under-resourced oversight of private regulatory functions.
More successful precedents include the global network of product safety certification bodies (such as UL and TÜV) that test products against government-mandated standards, and professional licensing regimes where private bar associations and medical boards enforce competence standards set by legislatures.