An IPCC for AI: the Scientific Panel and the Global Dialogue

Lesson 2 of 4 in The United Nations and the Treaty Track: Global Digital Compact and the CoE Convention.

On 26 August 2025, resolution A/RES/79/325 converted the GDC’s promise into institutions: the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance. The design borrows deliberately from climate governance — the IPCC model: separate the scientific assessment (what does the evidence say?) from the political negotiation (what should states do?), so that the science can be credible precisely because it is not negotiating anything.

The Panel is the scientific half. Forty members, drawn from all five UN regional groups, serving three-year terms — selected from over 2,600 applicants from more than 140 countries. The General Assembly appointed the slate on 12 February 2026; unusually for this track, the vote was not consensus — the United States and Paraguay voted against, a data point worth remembering when you assess US posture toward multilateral AI governance. At its inaugural meeting on 3 March 2026, the Panel elected its co-chairs: Yoshua Bengio, the Turing-award deep-learning pioneer who also chaired the International AI Safety Report, and Maria Ressa, the Nobel-Peace-Prize-winning journalist — one chair from the science of AI, one from its consequences for information and democracy.

Read the Panel’s mandate carefully, because its limits are the design: it issues an annual report that is policy-relevant but non-prescriptive — it tells governments what the evidence shows, never what to do. Its scope excludes military AI, which stays on the separate Geneva track (the CCW discussions on autonomous weapons). And it commands no budget or enforcement — like the IPCC, its only weapon is authoritative shared facts, which is exactly the weapon the fragmented AI debate has lacked.

The Global Dialogue is the political half: the standing forum where every UN member state — not 38, not 7 — discusses AI governance. Its launch was set for Geneva in 2026, with a second session in New York in 2027. The division of labour matters: the Panel produces the shared evidence base; the Dialogue is where states argue about what follows from it.

Key terms: scientific panel, global dialogue, ipcc model, unga resolution

Two other UN threads run alongside, and interviews love them. The ITU’s AI for Good summit has convened annually in Geneva since 2017 — older than the whole governance wave, focused on AI for the Sustainable Development Goals rather than on rules. And autonomous weapons are discussed by a Group of Governmental Experts under the Convention on Certain Conventional Weapons (CCW) — a deliberately separate, slower, consensus-bound track. When someone says "the UN is regulating military AI", the accurate answer is: it has been discussing it in Geneva for a decade, and the new Scientific Panel is explicitly fenced off from the topic.

Interactive checkpoint quiz (2 questions) — open this page in a browser to take it.