How can AI companies be regulated without hindering innovation?
Regulate AI through multiple competing jurisdictions with clear, enforceable rules rather than comprehensive global frameworks.
Lee Kuan Yew shows predictable rules attract more investment than perfect ones. Ostrom demonstrates overlapping authorities govern complex systems better than single regulators. Hayek proves markets discover AI risks faster than bureaucrats can anticipate them. Keynes argues tail risks require insurance mechanisms that markets underprice.
The council splits on intervention level but agrees on institutional design.
Confidence summary: High confidence on institutional design principles, moderate confidence on specific enforcement mechanisms.
1. The core argument
Singapore attracted global finance not by copying London's rules, but by making its own rules predictable and enforceable. This pattern holds for AI governance: companies will migrate to jurisdictions that combine regulatory clarity with institutional competence, not those promising the most innovation-friendly policies. The winners will be governments that regulate precisely what they can actually enforce.
The internet's early success occurred under regulatory ambiguity, but AI presents tail risks that markets systematically underprice. A social media algorithm that destabilizes elections imposes costs never captured in corporate profit statements. Yet central authorities cannot possess the distributed knowledge required to write effective rules for technologies they do not understand, applied to use cases they cannot foresee. The knowledge problem applies with special force to innovation regulation.
The solution lies in jurisdictional diversity with low exit costs. Multiple competing approaches allow policy learning while preserving reversibility. Estonia's framework will differ from California's, which will differ from Singapore's. This diversity enables discovery of what works without betting everything on a single regulatory architecture.
2. How each member frames it
Lee Kuan Yew sees regulatory competition as the mechanism that separates competent institutions from good intentions. Markets punish weak enforcement faster than they reward liberal policies.
Elinor Ostrom reframes the challenge through polycentric governance. Complex systems require overlapping authorities with clear boundaries, not single hierarchical frameworks that guarantee either capture or paralysis.
Friedrich Hayek views any comprehensive AI regulation as a knowledge problem. Central authorities cannot anticipate the discoveries that drive innovation or the risks that emerge from unforeseeable applications.
John Maynard Keynes emphasizes genuine uncertainty about AI impacts. Under such conditions, policy should insure against worst-case scenarios rather than optimize for expected outcomes.
Albert Hirschman champions regulatory ambiguity as productive ignorance. Initial uncertainty allows experimentation that would be impossible under rigid frameworks designed for imaginary perfect foresight.
3. Where the council agrees
The most surprising convergence appears around regulatory uncertainty itself. Even Hayek acknowledges that unclear rules can be worse than imperfect ones, while Hirschman argues that some ambiguity enables beneficial experimentation. All members agree that single comprehensive frameworks will either paralyze innovation or miss genuine risks.
They converge on the superiority of multiple experimental approaches over unified global standards. Singapore's success with financial services regulation came from adapting international frameworks to local institutional capacity, not from wholesale adoption. The same principle applies to AI: effective governance requires matching regulatory ambition to enforcement capability.
The council also agrees that exit options prevent regulatory capture and enable policy learning. Jurisdictions that over-regulate will lose investment and talent. Those that under-regulate will face crises they cannot manage. This competitive pressure improves regulatory quality more reliably than technocratic design.
Finally, they accept that innovation speed and safety margins cannot both be maximized simultaneously. The trade-off is real, and different jurisdictions will reasonably choose different points on this spectrum.
4. What would change this verdict
Evidence that AI development creates irreversible global risks requiring coordinated intervention would strengthen the case for comprehensive frameworks. Discovery that regulatory competition produces a race to the bottom rather than learning would favor harmonization efforts.