The General Assembly appointed forty experts to the Independent International Scientific Panel on Artificial Intelligence, the first standing scientific body the United Nations has created for the technology. The panel members serve a single term running from 12 February 2026 to 11 February 2029, and they were drawn from a field of more than 2,600 candidates.
The vote was recorded at the request of the United States. It carried 117 to 2, with Paraguay and the United States against and Tunisia and Ukraine abstaining. Ukraine's abstention was tied to the inclusion of Russian expert Andrei Neznamov on the roster. The panel includes Nobel Peace Prize laureate Maria Ressa, Turkish physicist Melahat Bilge Demirkoz, two United States academics and two Chinese representatives.
Secretary-General Antonio Guterres framed the case for the panel in terms of asymmetry between what any single government can see and what the technology is doing. "AI is moving at the speed of light. No country can see the full picture alone," he said.
The American position was equally direct. United States counselor Lauren Lovelace told the Assembly, "We will not cede authority over AI to international bodies influenced by authoritarian regimes." That leaves the largest concentration of frontier compute and frontier labs formally outside a body whose whole premise is a shared evidence base.
The panel's job is assessment rather than rulemaking. It is charged with tracking technical progress and the economic, ethical and social implications that follow, and delivering impartial analysis that member states can use. It has no enforcement power and issues no binding findings.
The companion process is the Global Dialogue on AI Governance, which held its first meeting in Geneva in July 2026 and is scheduled to convene annually alongside the AI for Good Global Summit. Its agenda covers safe development of AI systems, capacity gaps in developing countries, interoperability between national governance regimes, and socioeconomic effects. The panel's first report was expected in time for that July meeting.
Diplomatic activity continued through September. On 10 September 2026 the permanent missions of El Salvador and Estonia, co-chairs of the Global Dialogue, hosted an ambassadorial lunch on multilateral AI governance, positioning the next round of discussions.
What the structure produces is a split. Assessment is multilateral and now permanent. Regulation stays national, and the three jurisdictions that matter most for frontier capability are each building their own regime on their own schedule: the European Union through the AI Act, China through sectoral measures from its cyberspace regulator, and the United States through executive action that leans toward acceleration rather than restraint.
For anyone deploying these systems across borders, the practical read is that the UN panel will supply a common vocabulary and a shared factual record, and very little else. Compliance obligations will keep arriving from capitals, not from New York.