The United Nations Independent International Scientific Panel on Artificial Intelligence delivered its first preliminary report to all 193 member states on Wednesday, concluding that the technology is advancing faster than science can characterize it and faster than governments can meaningfully regulate it. The document, distributed ahead of the inaugural U.N. Global Dialogue on A.I. Governance in Geneva on July 6–7, arrives as the diplomatic architecture around A.I. is still being sketched, and it lands with the weight of a body designed to be difficult to ignore.

The panel itself is the story as much as its findings. Established by General Assembly Resolution A/RES/79/325 in August 2025 and convened in February 2026, its 40 members were selected from more than 2,600 applications spanning 140 countries, serving three-year terms. It’s co-chaired by Yoshua Bengio, the Turing Award-winning professor at the Université de Montréal, and Maria Ressa, the Nobel Peace Prize laureate and co-founder of Rappler. The pairing is deliberate: a foundational computer scientist and a journalist who has spent a decade documenting what algorithmic systems do to democratic societies.

Their language at the launch wasn’t diplomatic. “A.I. capabilities are outpacing both scientific understanding and governments’ ability to adapt,” Bengio said, adding that “with growing evidence of deceptive A.I. behaviour, science currently cannot guarantee that as capabilities continue to increase, A.I. will not cause catastrophic harm, either on its own or due to malicious users.” Ressa was blunter still. “The pace is not slowing; the power is concentrating; and control is not guaranteed,” she said. “no expert today can promise you that the most advanced systems will do what you instruct it to do.”

The report’s structural findings sharpen the warning. Roughly 75 percent of the computing power behind the world’s leading A.I. supercomputers sits in the United States, about 15 percent in China, with the two countries commanding around 90 percent combined. U.S. and Chinese firms are developing nearly all leading general-purpose models. More than one billion people now use conversational A.I. each week, though adoption lags sharply in developing countries. The panel also flags that sycophantic A.I. behavior has already been linked to several severe mental-health incidents, including documented deaths.

Secretary-General António Guterres framed the political stakes directly. “The science is here. We can no longer say we did not know. What we do with it is now up to all of us.” The more A.I. advances without shared rules, he warned, “the less say governments and people will have in the outcome.”

The panel is advisory, not regulatory. That’s the whole point, and the whole problem. The IPCC produced decades of settled science before governments moved; the question in Geneva next week is whether A.I. governance has decades to spare.

Sources