The UN Independent International Scientific Panel on AI

What the Panel is

The Independent International Scientific Panel on Artificial Intelligence is the first global scientific body on AI. It was established within the United Nations by the General Assembly through resolution A/RES/79/325, adopted by consensus on 26 August 2025. Its purpose is to give Member States and other stakeholders a shared evidence base on a technology that moves faster than the institutions governing it.

The comparison most often reached for is the Intergovernmental Panel on Climate Change, and it is useful up to a point: both are scientific assessment bodies rather than rule-making ones. The Panel itself is explicit that it is not a regulatory body. It does not set rules, enforce standards or prescribe policy. Its own description of its output is "policy-relevant but non-prescriptive".

How it came about

The Panel did not appear on its own. It was a commitment in the Global Digital Compact, adopted in 2024 as part of the Pact for the Future at the Summit of the Future, and it builds on the recommendations of the Secretary-General's High-Level Advisory Body on Artificial Intelligence.

Resolution A/RES/79/325 created two things at once: the Panel, and the Global Dialogue on AI Governance, a universal forum intended to bring all countries into the same conversation. The two are designed to work together. The Panel produces the assessment; the Dialogue is where it is presented and discussed. Reading either in isolation gives a misleading picture of what the United Nations built.

Who sits on it, and how they were chosen

The Panel has forty members drawn from all five United Nations regions, nineteen women and twenty-one men. They come from academia, the private sector, civil society and the technical community, and their expertise spans three bands: core technical AI, applied AI together with safety and infrastructure, and AI policy, ethics and impact.

Selection ran through an open call posted from 25 September to 31 October 2025, which drew more than 2,600 applications from over 140 countries. Candidates were assessed on merit against criteria of outstanding expertise, with interdisciplinary perspective, geographical and gender balance, and representation from developing as well as developed countries taken into account. The Secretary-General then recommended forty candidates and the General Assembly appointed them in February 2026.

At the inaugural meeting on 3 March 2026 the members elected two co-chairs: Yoshua Bengio of Canada, Professor at the Université de Montréal, Co-President of LawZero, founder of Mila and a recipient of the 2018 Turing Award; and Maria Ressa of the Philippines, a Nobel Peace Prize laureate, journalist and co-founder of Rappler. The full membership is published on the Panel members page.

What it can and cannot do

The mandate is narrower than the attention around it suggests, and the boundaries are worth stating plainly because they are widely misread.

It can produce an annual report of evidence-based scientific assessments on the opportunities, risks and impacts of AI, presented at the Global Dialogue on AI Governance. It may issue thematic briefs on fast-moving issues, update the General Assembly up to twice a year, form working groups, and consult informally with external experts.

It cannot regulate. It sets no rules, enforces no standards and prescribes no policy. Its scope is also limited to the non-military domain, which excludes a substantial part of the AI risk conversation from its remit by design.

One structural detail matters for how much weight to give its findings: the Panel's Secretariat is coordinated by the Under-Secretary-General and Special Envoy for Digital and Emerging Technologies, Amandeep Gill, through the UN Office for Digital and Emerging Technologies, with members from ITU and UNESCO. That Secretariat is enabling only. It does not direct the scientific work, and the Panel's findings are not subject to United Nations review or approval. The independence in the name is a real design constraint, not a courtesy.

What it has published

The Panel's first output, the Preliminary Report of the Independent International Scientific Panel on AI: Evidence-based assessment of opportunities, risks and impacts of AI, was released on 1 July 2026. It synthesises existing research rather than presenting new findings, and it was written to inform the inaugural Global Dialogue on AI Governance held in Geneva on 6 and 7 July 2026.

Treating it as a settled scientific position would be a mistake. The Panel describes it as the beginning of a sustained process and as reflecting the best available evidence at the time of publication. The report and the launch materials are on the Panel's resources page.

What comes next

The Panel has said it will deepen its evidence base through structured consultation with the scientific community and engagement with external experts, and that it will issue thematic briefs on emerging issues. The topics it has named include AI and the environment, AI and child safety, and AI governance instruments together with the evaluation of their effectiveness. It has also flagged sectoral briefs on AI in space, quantum, legal and judicial systems, and financial markets.

The annual report presented to the Global Dialogue is the fixed point in the cycle. Everything else is designed to move faster than an annual rhythm allows, which is a sensible answer to the problem that an annual assessment of a technology this fast is out of date on publication.

What it means if you have to govern AI in practice

For anyone running or assuring AI inside an institution, the practical significance is not that a new rulebook exists. It is that a citable, non-partisan evidence base now exists, produced by a body no single country or vendor controls. That is useful in exactly the situation where an internal argument stalls because two parties are citing incompatible industry sources.

What the Panel does not do is tell an organisation what controls to run. That gap is filled by frameworks built for implementation rather than assessment, and the mapping between the two is the useful work. The NIST AI Risk Management Framework remains the most practical structure for organising controls, ISO/IEC 42001 the route to a certifiable management system, and the independent examination of either sits with IT audit of AI systems.

The institutional context around this Panel, including how AI governance actually operates inside United Nations organisations, is covered on AI governance in the United Nations. The public-sector view is on public sector AI governance, and the regional picture, where the Panel's emphasis on representation from developing countries matters most, is on AI governance in the Arab region.

Common questions