A hundred experts, thirty nations, one report
The 2026 International AI Safety Report, led by Yoshua Bengio with input from more than 100 experts nominated by over 30 countries, is the broadest multilateral assessment of AI risk attempted. Breadth like that buys legitimacy and costs sharpness.
The 2026 International AI Safety Report, chaired by Turing Award recipient Yoshua Bengio and drawing on more than 100 experts nominated by over 30 nations, is the broadest multilateral assessment of AI risk produced to date.
What multilateral breadth buys
Standing. A finding endorsed across thirty nations is harder for any single government to dismiss as another country's anxiety, and it gives national regulators something to cite that is not a lab's own publication. For a field where most authoritative technical writing comes from the companies being regulated, an independent reference point is worth a great deal.
What it costs
Precision. A document that must survive review by a hundred experts across thirty nations converges on what everybody can accept, and what everybody can accept is rarely the sharp claim. Consensus documents are strongest on describing the landscape and weakest on saying which risk to act on first — which is the question a policymaker actually has.
They are also slow. A report reflecting a hundred contributors describes the state of knowledge some months before publication, in a field where eleven models shipped in twenty days.
How to read it
As a floor, not a summary. What survives that process is what is no longer seriously contested — which makes disagreement with it expensive, and agreement with it cheap. The frontier of the argument is elsewhere, in the papers the report cites.
AI Security Institute — Research & Publications → · Future of Life Institute — AI Safety Index — Summer 2026 → · Oxford AIGI — Legal Alignment for Safe and Ethical AI →