Yoshua Bengio, Founder of Mila, has co-chaired the first Preliminary Report of the UN Independent International Scientific Panel on AI, released on July 1. Bengio led the effort alongside journalist and Nobel Peace Prize laureate Maria Ressa, co-founder of Rappler. Several members of the Mila community also contributed to its creation.
The Institute’s involvement at various levels in this global initiative mandated by the UN is a concrete illustration of Mila’s commitment to applying scientific rigor for the common good and to ensuring that independent research plays a decisive role in the global governance of AI as it takes shape.
Established the previous year by UN General Assembly resolution, the Panel’s mandate is to advance scientific understanding of AI and ensure that international deliberations are informed by the best available evidence. This preliminary report marks the beginning of an ongoing body of work: a standing mechanism that will produce independent, science-based assessments of AI's capabilities, opportunities and risks over time. It was developed by 40 experts drawn from 37 countries, working in their personal capacity rather than on behalf of any government, company or institution.
A Warning on the Pace of Safeguards
One of the Panel's central findings is that current safeguards cannot keep pace with the growth of AI's capabilities, and that public policymakers cannot govern what they do not understand. At the launch press conference of the preliminary report in New York, UN Secretary-General António Guterres told reporters that scientific data is now available and that leaders can no longer claim otherwise, adding that from now on, everyone must be part of the solution. .
Bengio and Ressa were also careful to define what the report is not. It does not offer policy recommendations, a choice meant to protect the Panel's ability to stay strictly evidence-based rather than risk having its work become politicized. The report is also explicit about its limits. It does not address military applications of AI or lethal autonomous weapons systems, and it flags several areas, including AI's macroeconomic effects and its environmental footprint, where the evidence base is still too uneven to support firm conclusions. The co-chairs described the report instead as a common baseline, a body of evidence that governments everywhere, not just the handful of countries where AI is developed, can draw on to make informed decisions.
What Comes Next
The Panel's findings will be presented to governments and the public this week at the inaugural UN Global Dialogue on AI Governance in Geneva, running July 6–7, giving delegations a shared scientific starting point for their discussions. The Panel will continue building its evidence base through consultations and engagement with the scientific community. Its next annual report will inform the second Global Dialogue, to be held in New York in May 2027.
Read the full Preliminary Report on the UN Independent International Scientific Panel on AI website.
