Trump가 Xi Jinping과의 예정된 회의에서 AI 개발 위험을 다루도록 촉구받고 있으며, 냉전 핵 군비 협상과의 유사성이 지적되고 있다.

Senator Chris Coons wants President Trump to put artificial intelligence at the top of the agenda when he next meets with Chinese President Xi Jinping. The Delaware Democrat has framed the stakes in terms traditionally reserved for nuclear weapons: he compared the need for AI cooperation to the nuclear arms negotiations that defined Cold War diplomacy.

Speaking during a Bloomberg interview on September 13, Coons argued that the rapid pace of AI development has left legislators scrambling to keep up. Unilateral pauses on AI research, he said, are unrealistic given the competitive dynamics between Washington and Beijing. The only viable path forward, in his view, is a framework both countries agree to.

Coons bases his argument partly on production capacity. The US currently produces over 10 million advanced AI chips annually, while China manufactures roughly 200,000—a 50-to-1 ratio that Coons wants to preserve. He has championed the SAFE Chips Act, legislation designed to codify restrictions on exporting advanced AI semiconductors to China. The bill would formalize what has been a patchwork of executive-branch export controls targeting high-end processors like Nvidia's H-series GPUs.

US administrations across both parties have moved to limit China's access to cutting-edge chip technology, but those restrictions have largely been executive actions—subject to reversal or non-enforcement by a successor. Coons seeks to embed these restrictions in statute, making them far harder to overturn.

This isn't the first time AI has surfaced in Trump-Xi diplomacy. The two leaders met in Beijing in May 2026, where AI guardrails and exports of Nvidia's H200 chips were reportedly discussed. They departed without signing any formal agreement.

China has publicly stated its goal of becoming the world's leading AI power by 2030, which is why Coons invoked the nuclear comparison. The US and Soviet Union negotiated arms reduction treaties despite being existential adversaries, based on the logic that certain risks were too catastrophic for either side to manage alone. Coons believes AI belongs in that category.

Coons is not alone in pushing for a bilateral framework. OpenAI CEO Sam Altman suggested on the same day that a US-China AI safety treaty—even one as simple as a single page—could be worthy of a Nobel Peace Prize.

The obstacle remains enforcement. Nuclear arms treaties relied on extensive verification mechanisms: inspectors, satellite monitoring, warhead counts. AI doesn't lend itself to such physical auditing. Any meaningful AI safety treaty would need to solve the verification problem—something neither country has yet proposed a credible mechanism for.