Regulation needs plumbing as well as principles: establishing registrations for frontier models and autonomous agents and creating markets for private AI regulatory services would give governments practical tools to scale oversight and steer safe innovation.

Main Finding

Legal and regulatory infrastructure — not just substantive rules — is crucial for effective AI governance. Policymakers should prioritize building systems (e.g., registration regimes, identification schemes, and regulatory markets) that can generate, implement, and enforce rules for transformative AI.

Key Points

• Law’s role includes establishing mechanisms and institutions that make rules actionable, measurable, and enforceable, not only specifying the rules themselves.
• Three concrete infrastructure proposals discussed:
• Registration regimes for frontier models (records of model capabilities, ownership, or deployment status).
• Registration and identification regimes for autonomous agents (ways to track, attribute, and identify agentic systems in deployment).
• Regulatory markets that allow private companies to innovate and deliver regulatory services (contracted or market-based intermediaries that help implement oversight and compliance).
• Emphasis on the transformative character of AI: rapidly changing capabilities raise unique needs for flexible, scalable institutional arrangements.
• The perspective is normative and design-focused: it argues for infrastructure choices that make future regulations practicable.

Data & Methods

• Type: Perspective / policy review and conceptual analysis.
• Methods: Argumentative review using illustrative examples and institutional design reasoning rather than empirical estimation or novel data analysis.
• No primary datasets or econometric methods are reported; conclusions are driven by policy logic, institutional comparisons, and forward-looking design considerations.

Implications for AI Economics

• Incentives and investment:
  • Registration and ID requirements will change firms’ marginal costs (compliance, disclosure) and may alter R&D and deployment incentives for frontier models and agentic systems.
  • Regulatory markets can create new revenue streams and business models (compliance-as-a-service), shifting where value accrues in the AI ecosystem.
• Market structure and competition:
  • Compliance burdens could favor larger incumbents with scale advantages, but well-designed regulatory markets might lower entry costs by outsourcing compliance to specialists.
  • Information revealed via registries reduces information asymmetries, affecting pricing, contracting, and risk-sharing in AI-related markets.
• Externalities and social welfare:
  • Better identification and registries improve externality internalization (e.g., liability, monitoring of harms), potentially lowering systemic risks and negative spillovers.
  • Trade-offs exist between transparency (reducing harms) and incentives for innovation/confidentiality.
• Enforcement and monitoring economics:
  • Registries and IDs reduce detection and attribution costs, changing optimal enforcement strategies and fine structures.
  • Market-based regulatory providers create principal–agent problems and potential regulatory capture that require contract and incentive design.
• Research opportunities:
  • Quantify compliance costs and their distribution across firms of different sizes.
  • Market design for regulatory services: competition, quality assurance, and moral hazard mitigation.
  • Welfare analysis of registries/IDs: benefits from reduced information frictions vs. costs to innovation and privacy.
  • International coordination externalities and cross-border enforcement mechanisms.
  • Dynamic modeling of how infrastructure affects innovation speed, diffusion, and systemic risk.