UN's AI Strategy Overlooks Labs That Build It
The United Nations' new AI strategy focuses on its use and governance, but leaves the development and evaluation of frontier AI systems outside the organization's direct purview.
The Geneva Digital Week opened July 6 with the Global Dialogue on AI Governance and the AI for Good Global Summit, where the UN's new Independent International Scientific Panel on Artificial Intelligence presented its first global scientific assessment of AI to governments.
This gathering caps three years of UN work on AI, focusing heavily on the demand side: harnessing AI for sustainable development goals, monitoring societal effects, and establishing ethical guardrails. However, the supply side—where frontier AI is produced, evaluated, and released—lacks meaningful UN presence.
There is no multilateral body with technical staff to examine lab work, nor a shared infrastructure for cross-border incident reporting or evaluating training runs. The governing architecture is currently consolidating through bilateral arrangements between frontier labs and hosting governments, often influenced by private entities and export controls.
This leaves the authority to determine access to frontier AI technologies residing by default in single national administrations, leading to frustration among other nations. The UN has precedent for this type of technical oversight, such as the International Atomic Energy Agency (IAEA), which monitors nuclear material production.
The UN's UN80 reform agenda presents an opportunity to address this gap. Establishing a small, standing evaluation unit using existing UN assets could invite labs to extend their voluntary agreements beyond current bilateral partners. This would grant labs global credibility, protect them from regulatory fragmentation, and ensure UN engagement extends to where AI is produced, not just where its consequences are felt.