The 40 member Independent International Scientific Panel on AI, co chaired by Nobel laureate Maria Ressa and Turing Award winner Yoshua Bengio, says governments cannot keep pace with the technology — and neither can science itself.
The United Nations has built a scientific body for artificial intelligence that no government or company can build alone. Its first message to the world is that governments cannot keep up.
The 40-member Independent International Scientific Panel on AI released its preliminary report on Wednesday, the first product of a body created to give policymakers something they currently lack: a shared, rigorous evidence base on what AI can do, what it is already doing to people, and where the limits of current knowledge actually sit. Co-chaired by Nobel Peace Prize laureate Maria Ressa, the Rappler CEO and co-founder, and Turing Award-winning AI researcher Yoshua Bengio, the Panel frames itself as the scientific common denominator that fragmented national regulators and the AI industry itself both need.
"The technology is transformative, but if the world keeps moving along this trajectory, humanity will fail to realize the gains it promises," Ressa said in the Panel's statement. "The risks, to societies, to security, and to our species, are too high, and the forces driving AI forward are not the forces that will deliver its benefits."
That tension, between AI's rapid deployment and the institutions trying to oversee it, is the Panel's central claim. AI capabilities are moving faster than scientific understanding and faster than governments' ability to adapt, Bengio said in the same statement, and "with growing evidence of deceptive AI behavior, science currently cannot guarantee that as capabilities continue to increase, AI will not cause catastrophic harm, either on its own or due to malicious users."
The Panel's risk taxonomy is unusually specific for a UN body. The report names deceptive AI behavior, where systems learn to mislead users or oversight mechanisms; mental-health harms to users, especially young users; the use of AI as a destructive tool; broader impacts on social, economic, and environmental systems; and what it calls challenges associated with controlling the technology, meaning the open question of whether increasingly capable systems can be reliably constrained once deployed.
The opportunity side of the same report is the counterweight the Panel is explicitly trying to keep on the table. "Deployed and applied thoughtfully, AI can support progress towards achieving the Sustainable Development Goals," the executive summary says, referring to the UN's 17 targets for global development by 2030, from poverty reduction to clean energy. The Panel also points to AI advancing health science and broadening access to education, framing both as genuine gains that the same governance gap could foreclose.
The Panel itself was set up to fill a recognized hole in global AI governance. Its mandate is narrower than the UN Secretary-General's AI Advisory Body, a separate high-level body that has focused on principles and recommendations; the new Panel is meant to be the rigorous scientific underpinning those policy efforts have lacked. The Panel's work lands days before the Global Dialogue on AI Governance in Geneva on 6-7 July, organized by UNESCO, and feeds into broader UN General Assembly processes later this year.
The interesting governance question is what an advisory body without enforcement power can actually do. The Panel's leverage is structural rather than legal. National regulators, the major AI companies, and downstream multilateral institutions all need some common reference point for what counts as AI risk, what the scientific consensus is, and where the unresolved questions lie. By setting itself up as that reference point, the Panel gains the kind of authority that does not require a treaty. Industry and civil society coverage describes the Panel as a direct answer to the absence of a globally shared scientific baseline.
Whether that authority holds depends on whether the Panel can stay scientifically credible as the industry races ahead and as governments push back on inconvenient findings. The Panel's own epistemic caveat is the most important sentence in the report. "Science currently cannot guarantee" safety as capabilities grow. That is not a hedge. It is an admission that the scientific community is being asked to govern a technology whose risks it does not yet fully understand, and that the next decade of AI governance will be built on a foundation that is still being laid.
What to watch: the Geneva dialogue on 6-7 July, where the Panel's preliminary findings are expected to be tested against positions from major governments and industry; the Panel's full report, due after the dialogue; and whether national AI regulators start treating the Panel's assessments as their working baseline or as one input among many.