Co-chairs of the United Nations’ Independent Scientific Panel on Artificial Intelligence said at a UN briefing that power over AI development is concentrated in a handful of companies and nations, a pattern that threatens equity, national sovereignty and the ability of many countries to shape governance.
Maria Resa said one benchmark of concentration is compute capacity: "The United States alone accounts for 75% of the computing power in the world's largest AI clusters," she said, adding that 91% of notable models in 2025 came from the private sector. The panel highlighted choke points across the technology stack, including lithography, fabrication, and chip design.
Yoshua Benjou told reporters the report is global and not tailored to any single country, but he said some countries—India was discussed as a relatively advantaged case because of its talent pool—still face constraints if they lack testing capacity and access to compute and models. "Individually it's clear they can't do anything," Benjou said of many smaller countries; the panel argued international coordination and capacity building are required to avoid an expanding divide.
In question-and-answer exchanges, journalists asked what concrete commitments the panel would press for at the Geneva dialogue; co-chairs reiterated the panel's evidence mandate and said policy prescriptions are for member states but that the report documents levers such as independent assessment, technical secondments for regulators, and incident reporting frameworks.
The co-chairs said the evidence dilemma—policymakers need evidence but evidence lags development—means that standardized reporting, independent evaluators with privacy-preserving access, and capacity investments outside a handful of high-capacity countries are immediate priorities for addressing the divide.
The panel will share the report publicly and bring its evidence to the Geneva dialogue on July 6–7.