The UN Can Diagnose AI's Risks. It Cannot Coordinate the Powers That Create Them.

The UN was built to coordinate when nations won't. In AI governance, it has become the place where the failure to coordinate is most visible. Its Independent International Scientific Panel on AI, convened in Madrid in April, warned that the US and China together control 90% of the world's AI computing power ^[1]. The head of the International Telecommunication Union, the UN agency overseeing the effort, cautioned about a "second great divergence" between nations that produce AI and nations that consume it ^[2]. At the Geneva Global Dialogue on July 6, Secretary-General Guterres called for guardrails, a Global Fund for AI, a ban on lethal autonomous weapons, and a child-safety pledge ^[3]. None of it was binding. The panel's mandate is to inform the dialogue, not to govern it ^[2]. The summit's concrete institutional output was scheduling a second meeting in New York next year. When Guterres flagged AI-enabled terrorism at a counter-terrorism conference days earlier, his Acting Under-Secretary-General, Vladimir Zouev, made the deferral explicit.

The primary responsibility for preventing and combating terrorism falls on member states. — Alexandre Zouev

The UN names the risk. It proposes. Then it hands the problem back to the same national governments where fragmentation already exists. This deferral would matter less if those governments were coordinating among themselves. They are doing the opposite. The US launched Pax Silica, a 35-nation AI supply-chain initiative explicitly defined against China ^[4]. US officials frame the competition in stark terms.

We’re an AI superpower. — Scott Bessent

China is building its own parallel architecture. It organized a 65-nation coalition within the UN Human Rights Council to push a "people-centered, development-oriented" AI governance framework ^[5]. It announced a rival World AI Cooperation Organization in its Global Governance Initiative white paper ^[6]. Its top diplomat, Wang Yi, toured Finland and Sweden in July building bilateral AI ties ^[7]. And through all of this, China's UN Ambassador Fu Cong affirmed support for the UN's "central role" in AI governance ^[8]. The rhetorical affirmation and the parallel institution-building are complementary: occupy the multilateral space rhetorically while building competitive blocs within it. The UN's "central role" becomes the venue where fragmentation happens, not the mechanism that overcomes it. The G7 summit at Évian-les-Bains in June tested whether a smaller group of like-minded nations could do better. They agreed only on child safety: safety-by-design, age verification, zero-tolerance for AI-generated deepfakes of minors ^[9]. On the contentious question of who gets access to advanced AI models, the summit produced no resolution. French President Macron warned of what failure would mean.

we will fracture the world, and we will have no effective solution — Emmanuel Macron

The governance vacuum is now acute enough that a private company stepped into it. Anthropic proposed a nuclear-arms-control-style pause and verification regime for frontier AI systems ^[10]. Its competitor OpenAI rejected the idea, saying democratic governments, not private companies, must determine the rules. No government adopted the proposal. The pattern runs from the bottom of the system to the top. The Justice Department sued to block Colorado's state AI law ^[11]. The EU failed to soften its own AI Act after twelve hours of negotiation ^[12]. India's Supreme Court drafted its own rules for judicial AI ^[13]. Each jurisdiction governs with the tools at hand. Nothing coordinates them. And at the top, the institution built to do that work can diagnose the problem but defer the action, convene the dialogue but not bind the outcome. The next UN meeting, scheduled for New York next year, will convene in a world where the US has organized 35 nations against China and China has organized 65 within the UN's own chambers. The institution designed to coordinate will diagnose the problem again. The powers that created it will keep pulling apart.