Global Coordination
Climate, pandemics and AI ignore borders, yet binding, enforceable agreements between states rarely succeed.
Open in the interactive tree →Common goods invite free riding (game theory), and the UN rests on consensus. Successes exist (Montreal Protocol 1987 on the ozone layer, Paris Agreement 2015), but they are the exception.
As of October 2026
Climate: COP30 in Belém (November 2025) acknowledged an overshoot of 1.5 degrees for the first time in a COP text but adopted no fossil-fuel roadmap; new national pledges from 119 countries cover only about 11% of the cuts needed by 2035 for 1.5 degrees. Plastics treaty talks failed in Geneva on 15 August 2025 and a further session in February 2026 was mainly administrative, while the PABS annex of the pandemic agreement was postponed to May 2027. For AI, the UN's 40-member Scientific Panel (co-chairs Bengio and Ressa) published a preliminary report on 1 July 2026 for the first Global Dialogue on AI Governance in Geneva on 6-7 July 2026, without binding effect. In tax, a 'side-by-side' package accommodating the US on the OECD minimum tax was agreed in January 2026 while UN talks on a tax framework convention continue.
What is missing
- Enforcement mechanisms and consequences for breaches (most commitments are voluntary)
- Fair burden-sharing between rich and poor states (finance, technology, access to vaccines and data)
- Decision procedures beyond unanimity, e.g. coalitions of the willing with trade incentives
- Independent shared measurement and verification (emissions, pathogens, AI capabilities)
- Political will in states with short election cycles
Becomes possible once solved
- Fast coordinated response to pandemics
- A binding path to climate neutrality
- Shared safety standards for advanced AI
- Reliable taxation of multinational companies
Open steps
- Independent verification of commitments High AI leverageBuild shared, independent measurement of emissions, pathogens and AI capabilities that states cannot dispute, so commitments can be checked.
- Enforcement that actually works Medium AI leverageFind which enforcement designs (trade incentives, conditional pledges, coalitions of the willing) raise compliance, using past treaties and models.
- Burden-sharing rules states accept Low AI leverageDevelop burden-sharing formulas for finance, technology and vaccine access that rich and poor states both accept, and test their acceptability.
- Consensus-finding at scale Medium AI leverageTest whether AI-assisted mediation helps delegations and citizens' assemblies find common ground faster and more fairly.
Where AI could help
Medium AI leverage. AI can help verify commitments and support consensus-finding; political will, enforcement and burden-sharing remain the core block.
- Verify emissions, methane leaks and other treaty compliance independently from satellite and sensor data
- Mediate and draft consensus texts for negotiators and citizens' assemblies
- Simulate treaty and enforcement designs with game-theoretic agent models before negotiating
- Summarize and translate huge negotiation texts so small delegations can keep up
Shown so far
- In October 2024 DeepMind's Habermas Machine (Science) wrote group statements that participants preferred over human mediators' statements 56% of the time, in UK deliberations with 5,734 participants. source
- In November 2022 Climate TRACE used AI on satellite and sensor data to publish emissions for 72,612 facilities, finding oil and gas emissions up to three times higher than self-reported by some top countries. source
Prerequisites
- Bretton Woods, IMF, World Bank1944IMF and World Bank are the standing bodies of global economic coordination
- Game Theory1944
- United Nations1945
- Nuclear Non-Proliferation Treaty1968A long-lived arms-control treaty, in force since 1970 with near-universal membership
- WHO Pandemic Agreement2025