The UN's first global AI scientific body says relying on overseas models and cloud infrastructure buys capability without governance authority, and the window to fix it is closing.
The first global scientific body on AI is telling governments that getting access to the technology and keeping control of it are now opposite projects. Countries can adopt frontier AI tools through foreign model providers and overseas cloud infrastructure without ever building the local capacity to govern how those systems behave, what data they train on, or which safeguards apply. That gap between access and authority is the mechanism through which AI is on track to widen global inequality, according to the UN Independent International Scientific Panel on AI's preliminary report released this week.
The panel was established by the UN General Assembly in 2025 and styled as the world's first global scientific body on AI. Its preliminary report, issued Wednesday, frames AI as a dual-use technology: capable of transformative gains in agriculture, education, and public services, and equally capable of catastrophic misuse through AI-enabled fraud and election influence operations. The report names the adoption pathway, not adoption itself, as the risk.
"Access to AI tools alone does not produce equal benefit," the panel's preliminary report states directly. Governments that rely on foreign models, cloud infrastructure, and data pipelines may gain working AI capability while losing practical control over the standards, safeguards, and local benefit capture that determine who actually wins from the technology.
Secretary-General António Guterres sharpened that point at a press conference the same day: "The more AI advances without shared rules, the less say governments and people will have in the outcome." His warning to governments was blunt: "Do not wait … the science is here. We can no longer say we did not know what we do."
The panel's proposed response is a shared governance framework, not a binding treaty. It calls on governments to coordinate before the asymmetry window closes, positioning the moment as a one-shot choice about who sets the rules AI will run on. Coverage in Euronews the next day carried the same message about the closing window.
Naming sovereignty as the issue changes the question. The generic inequality story asks whether the technology will concentrate gains. The panel asks instead which countries will be able to set the terms under which AI operates inside their borders, and which will be locked into dependency on providers whose incentives they do not shape.
The panel's report is preliminary. The asymmetry it names is interim, and the governance framework it proposes is non-binding. The next trigger is the panel's full report, expected later this year, which will test whether shared scientific attention translates into shared rules.