The global debate over artificial intelligence is shifting from voluntary safeguards to demands for international authority—just as the United States and China, the two most consequential AI powers, remain outside the coalition driving the push.

Led by Norway and Finland, 20 nations have backed a declaration calling for the United Nations to examine an institution capable of setting standards, verifying compliance and convening governments when AI systems cross dangerous capability thresholds. The signatories present the effort as a way to manage risks without allowing access to AI’s benefits to widen the divide between countries.

Australia’s Anthony Albanese framed the coalition’s position as pro-innovation but firmly bounded by human judgment. AI could “improve lives,” strengthen productivity and advance medical science, he said, but “humans need to be in control of it.”¹ The declaration similarly says AI must remain under “human direction, oversight and control” and operate within international law.²

The proposal also reflects concerns raised by leading technology executives, including Anthropic CEO Dario Amodei and OpenAI chief Sam Altman. Its supporters argue that common testing, evaluation and verification are necessary because advanced systems can create risks in cybersecurity, biology and human control that no single country can manage alone.

Yet the coalition’s apparent consensus masks a sharp geopolitical and political divide. The United States, China, Britain, France, Italy, Poland and Japan have not signed. Critics portray international oversight as a bid for expanded government control, a technology cartel favoring incumbent companies, or a fear-driven strategy that could slow the United States while China advances.

Even the proposal’s supporters acknowledge the central weakness: an institution may be feasible, but giving it “real teeth” and verifying secret or military AI systems would be difficult.² The declaration therefore represents less a functioning global regime than an opening claim in a contest over who will define AI’s limits—and whether the most powerful players will accept them.