The UN steps in: the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance

For two years the United Nations watched from the sidelines while the European Union, the Council of Europe and national legislators built the architecture of AI regulation. In 2026 it stepped in — not with a binding instrument, but with the two things it is uniquely placed to supply: an authoritative scientific voice and a negotiating table that seats every country, not only the technologically advanced ones.

One resolution, two new bodies

Both bodies rest on a single instrument: General Assembly resolution A/RES/79/325, adopted by consensus on 26 August 2025, itself a follow-on to the Global Digital Compact (A/RES/79/1) agreed at the 2024 Summit of the Future. The resolution created an Independent International Scientific Panel on AI and a Global Dialogue on AI Governance — deliberately modelled on the division of labour that has served climate policy: a scientific body to establish what is known, and a political forum to decide what to do about it.

The Scientific Panel: an IPCC for AI

The Panel is the UN's answer to a problem every regulator faces — the evidence base for AI risk is contested, fast-moving and dominated by the firms being regulated. It has 40 members, recommended by the Secretary-General from more than 2,600 applications across over 140 countries and appointed by the General Assembly in February 2026 for a three-year term. The membership is balanced across the five UN regions and by gender (19 women, 21 men). At its inaugural meeting in New York on 3 March 2026 the Panel elected two co-chairs whose names signal its intended weight: Yoshua Bengio, a founding figure of modern deep learning, and Maria Ressa, the Nobel-laureate journalist.

Its output is a single, deliberately constrained product: one evidence-based annual report on the opportunities, risks and impacts of AI. The constraint is the point. The Panel does not regulate, does not certify and does not adjudicate; it provides the independent scientific reference that the political process can draw on without having to trust any one government or company. The Secretary-General framed its mandate as "intentionally broad — spanning from frontier systems to the impacts already unfolding across societies and economies."

The Global Dialogue: the political forum

The Panel feeds the Global Dialogue on AI Governance, the intergovernmental forum where its findings meet negotiation. Its first substantive session is in Geneva on 6–7 July 2026, a two-day event with a high-level segment, thematic sessions and side events, and the Panel's first annual report as its reference point. The Dialogue's explicit purpose is distributional: to ensure AI governance "reflects priorities across all nations" and that AI's benefits are shared, "rather than being controlled by the most technologically advanced countries."

Why a non-binding layer still matters

Neither body issues rules, and it would be easy to dismiss them as talking shops. That misreads how international norms harden. The OECD AI Principles are also non-binding, yet the AI Act borrowed its very definition of an AI system from them; the G7 Hiroshima Process is voluntary, yet it now structures how frontier developers report. The UN layer matters for three reasons. First, legitimacy of evidence: a Panel report carries a universality no national institute can claim, and it becomes the citation that later instruments reach for. Second, convening power: the Global Dialogue is the one venue where states outside the EU–US–China triangle set the agenda, which shapes what eventually becomes treaty language. Third, interoperability: as jurisdictions from Vietnam to Brazil write their own AI laws, a shared scientific baseline is what keeps them from diverging into mutually unrecognisable regimes. For organisations, the practical reading is modest but real: the UN layer is where the next generation of binding rules will find its vocabulary and its evidence — the same path the OECD Principles travelled into the AI Act.