Surviving AI: Trump and Xi Jinping Negotiate the Terms of Rivalry, Not Global Governance

The AI Race Is Global. Its Rules Are Still Being Written

September 24, 2026

Chinese President Xi Jinping arrived at the White House on Thursday for his first state visit to Washington in more than a decade, greeted by President Donald Trump with the full machinery of a state occasion. The pageantry was impressive, but the substance is considerably harder. The two leaders are confronting a relationship that has moved beyond the old vocabulary of tariffs and trade deficits. Artificial intelligence, advanced semiconductors, critical minerals and national security are now intertwined, while Iran has added a live geopolitical crisis to an already dangerous relationship. Trump has said AI will be a “big topic” of the talks, but also indicated that he wants the existing situation left largely intact.

That may be the central contradiction of the summit. Washington and Beijing are competing to build increasingly powerful AI systems while simultaneously acknowledging that some of the consequences could become catastrophic. Treasury Secretary Scott Bessent has already proposed a U.S.-China notification mechanism for AI incidents that rise to the level of national security threats, along with a broader bilateral AI dialogue. The idea is modest: if an AI system produces an event capable of creating a national-security crisis, Washington and Beijing should at least have a telephone number to call each other. But that is crisis management, not global safety regulation. The two governments are still trying to win the race while discussing how not to crash into each other during it.

And the AI race is moving faster than diplomacy. The United States retains enormous advantages in AI investment and computing power, while China has been closing gaps in research and advanced models. Both governments regard AI as an instrument of economic and strategic power, making the prospect of genuinely slowing frontier development politically difficult. Yet something remarkable happened just as Xi arrived: the heads of OpenAI and Anthropic went before the U.N. Security Council and argued that AI needs international rules. Sam Altman warned that the future of AI cannot be determined by a handful of private laboratories, while Anthropic CEO Dario Amodei warned that poorly managed AI could become a risk to humanity. The U.S. administration, however, told the council that it rejects centralized international control of AI.

That creates an extraordinary diplomatic paradox. The companies building frontier AI are asking governments to establish rules; governments are competing to ensure that their own companies remain ahead; and the two countries with the greatest technological leverage are discussing only a limited notification mechanism. Meanwhile, 16 Democratic senators led by Tim Kaine have urged Trump to negotiate formal U.S.-China guardrails covering development, testing, monitoring and human control of autonomous systems, even proposing consideration of a mutual pause in frontier AI development. That proposal does not represent an agreement between Washington and Beijing, but it demonstrates how rapidly AI safety has moved from an abstract technology debate into a national-security issue.

Trade is the more immediate area where the summit may actually produce something tangible. On Wednesday, Bessent said Washington and Beijing had agreed to extend their existing trade truce for two months beyond its Nov. 10 expiration while negotiating a larger agreement. Critical minerals remain a major pressure point, as does the implementation of earlier promises involving U.S. agricultural exports, Boeing aircraft and reciprocal tariff reductions. Progress has been uneven: China has not yet delivered the promised 200 Boeing orders, and U.S. officials continue to complain about the flow of rare earths and other critical minerals. The difference from last year's tariff confrontation is significant, however. The two governments are now trying to stabilize the relationship without pretending that the underlying strategic competition has disappeared.

Iran is the uncomfortable third chair at the table. Trump wants Chinese influence over Tehran, but Beijing is deeply connected to Iran economically and remains the country's biggest buyer of Iranian oil. Washington has imposed sanctions on Chinese companies and smaller refineries involved in Iranian trade, while so far avoiding the most disruptive measures against major Chinese financial institutions. Reuters reports that expectations of Xi pressing Iran into submission are limited, precisely because Beijing has its own interests in maintaining the relationship. Iran therefore turns the summit into something larger than a bilateral negotiation: Washington is asking Beijing to help manage a war while simultaneously trying to contain Beijing's economic relationship with the country at the center of that war.

The most consequential question in Washington may therefore have nothing to do with tariffs, Boeing or even Iran. It is whether the United States and China can recognize that some technologies are too consequential to be governed entirely by competition. A notification hotline is useful; a trade truce is useful; diplomatic contact is indispensable. But none answers the harder question now being posed at the United Nations: who sets the rules when the technology being raced toward may eventually challenge the ability of governments themselves to control it? Xi and Trump can negotiate the terms of their rivalry. The rest of the world needs them to help establish the rules for surviving it.