Which global institutions should exist to prevent an arms race in dangerous technologies like AI?
Build overlapping institutions at different scales rather than one global AI authority.
Schmidt argues democratic allies must set binding standards first, then negotiate from strength. Roosevelt counters that excluding major powers destroys legitimacy and guarantees non-compliance. Lee Kuan Yew and Ostrom both see the solution in multiple frameworks: bilateral agreements on specific risks, alliance standards among democracies, regional protections for middle powers.
Sun Tzu adds that intelligence-sharing institutions must stay separate from development institutions to prevent miscalculation without stopping competition.
Confidence summary: Strong convergence on polycentric solutions, with moderate disagreement on whether democratic alliances should establish standards first or include all major powers from the start.
1. The core argument
Schmidt drove NATO's 1979 missile decision precisely because technology competition between adversaries cannot be managed by institutions that include the competitors. The lesson holds for AI: China views artificial intelligence as regime survival insurance, America as competitive advantage. No universal framework bridges that gap because the divide runs deeper than technical standards—it is about power itself.
Yet Roosevelt's counter cuts deeper. When she included Stalin's delegation in drafting the Universal Declaration of Rights, she understood that moral frameworks without universal participation become mere alignment exercises. The excluded powers race ahead while the included ones constrain themselves. Today's AI governance faces the same legitimacy trap: democratic-only institutions may produce elegant standards, but they cannot prevent arms races they do not include.
The synthesis emerges from Singapore's position between giants. Lee Kuan Yew learned that federal arrangements collapse when partners reject shared rules, but that middle powers suffer most when great power competition forces binary choices. AI governance must be polycentric—multiple overlapping institutions serving different functions rather than one universal authority managing everything.
2. How each member frames it
Helmut Schmidt sees alliance cohesion as the prerequisite for effective negotiation, arguing that democracies must build binding AI standards among themselves before engaging competitors from strength.
Eleanor Roosevelt reframes exclusion as the fundamental problem, insisting that universal participation provides the only legitimate foundation for constraining technology races between major powers.
Lee Kuan Yew views this through the lens of small state survival, recognizing that both universal and alliance approaches create impossible choices for countries caught between competing technology blocs.
Sun Tzu approaches AI governance as strategic competition management, focusing on intelligence-sharing mechanisms that prevent miscalculation without requiring cooperation on development itself.
Elinor Ostrom transforms the debate by identifying multiple overlapping commons problems that require matched institutional solutions at different scales and stakeholder levels.
3. Where the council agrees
The most surprising consensus emerges around institutional multiplicity. Even Roosevelt, the champion of universal frameworks, acknowledges that no single institution can manage competition across technical standards, deployment protocols, and strategic stability simultaneously. Schmidt's alliance-first approach and Sun Tzu's intelligence-sharing mechanisms serve different functions within the same polycentric system.
All members recognize that pure competition destroys shared infrastructure while pure cooperation ignores strategic realities. The solution lies in separating cooperative functions from competitive ones: safety research can be shared while deployment advantages remain protected. Technical communities can collaborate on standards while nation-states compete on applications.
Most critically, the council agrees that middle powers need protection from great power technology competition. Whether through regional frameworks, alliance structures, or universal institutions, smaller countries require mechanisms to avoid forced choices between American and Chinese AI systems. This shared vulnerability creates space for institutional cooperation even between strategic competitors.
4. What would change this verdict
Evidence that one major power is achieving decisive AI breakthrough would force faster institutional consolidation among competitors. Successful demonstration of universal AI safety cooperation would strengthen Roosevelt's case for inclusive frameworks over Schmidt's alliance approach.