Open problem 10
The International Institution Design Problem
- frontier AI
- transformative AI
- AGI
Should global AI governance be built through the UN, a new AI agency, standards bodies, minilateral clubs, treaty regimes, compute-governance institutions, scientific panels, or issue-specific networks?
Why it’s foundational
AI is transnational, but regulatory authority remains national. Governance needs legitimacy, speed, technical capacity, enforcement, and inclusion. No existing institution has all five.
Why it’s difficult
Global institutions are slow; minilateral clubs lack legitimacy; standards bodies lack enforcement; national-security institutions lack transparency; companies hold much of the relevant information.
Hidden assumptions
The field overuses analogies: IAEA, IPCC, ICAO, FATF, CERN, WTO. Each analogy highlights something useful and hides something crucial. AI is not simply nuclear, climate, aviation, finance, or science infrastructure.
Competing positions
- UN-centred inclusive governance
- G7/OECD-style club governance
- Treaty-based legalism
- Standards-led interoperability
- Regulator-to-regulator networks
- AI safety institutes
- Compute-governance regimes
- No global institution, only domestic law
What could make progress
Institutional design comparisons; stress tests against AGI timelines; treaty-verification analysis; mapping of which functions need universality versus technical depth; pilot institutions for information-sharing and joint evaluations. The UN’s current approach emphasises inclusive dialogue and scientific assessment, while the Council of Europe treaty focuses on human rights, democracy, and rule of law. The unresolved question is whether such institutions can govern frontier capability races, not merely articulate norms.
What it would change
It would determine whether international AI governance prioritises legitimacy, speed, enforceability, scientific assessment, capacity-building, or strategic restraint.
Sub-agenda
- Which AI governance functions must be global rather than domestic?
- What should be centralised, and what should remain networked?
- Can an international scientific panel influence high-stakes decisions in time?
- What enforcement powers, if any, should an AI institution have?
- How should private labs be represented without letting them dominate?
Priority (editor scoring)
Important, but often stuck in analogy rather than mechanism.
- Importance
- 4/5
- Neglected
- 3/5
- Difficulty
- 5/5
- Actionable
- 4/5
- Robust
- 4/5
- Nat’l+int’l
- 5/5
Where the catalog bears on this
Instruments this puzzle names that are in the catalog:
Editorial content — a human-authored agenda question, rendered verbatim. No part of this analysis is AI-generated (see the charter).