The AI Race Could Sink US-China Détente
The competition between the United States and China to develop AI and deploy it globally poses the biggest threat to the two countries’ currently constructive relations. Without an ongoing AI dialogue to avoid escalation, a return to the zero-sum conflict that rattled both sides last year cannot be ruled out.
Ian Bremmer, Founder and President of Eurasia Group and GZERO Media, is a member of the Executive Committee of the UN High-level Advisory Body on Artificial Intelligence.
__________________________________________–
NEW YORK—Just three months after US President Donald Trump returned to the White House in 2025, he launched a new trade war with China. It didn’t go well. Immediately after he announced a 34% duty on all Chinese imports, China signaled that it was ready to play hardball with a matching 34% tariff on all US goods and immediate export controls on seven rare-earth elements and the rare-earth magnets used in many 21st-century consumer products and military technologies.
After another quick round of back-and-forth retaliation, Trump backed down, and the two sides met in Geneva to declare a truce. Six months later, that truce was extended, and it remains in force today as Trump and President Xi Jinping prepare for their summit later this month.
Both sides have (for now) recognized that confrontation would badly damage both economies. Pragmatism became the new rule for the world’s most important bilateral relationship. US and Chinese officials have since looked for and found ways to work together more constructively on trade and investment, as well as engaging in more frequent, higher-level military-to-military communications. To keep relations on track, the Trump administration has delayed new defense packages for Taiwan, allowed restrictions on commerce with Hong Kong to lapse, and scaled back the periodic US naval patrols that are meant to assert freedom of navigation in the South China Sea.
The big question for a world now plagued by two major wars and another barrage of Trump-generated trade conflicts is how long these US-Chinese guardrails will hold. On a positive note, the current détente depends on a balance of coercive leverage that should last, in most areas, for at least the next several years.
The US will continue to reduce its direct reliance on Chinese goods imports, but it will remain vulnerable to China’s control over processed critical minerals and intermediate goods. China will continue to reduce its foreign dependencies and redirect exports toward the Global South, but it will still be vulnerable to US financial sanctions and some technology export controls.
To be sure, the US and Chinese leadership won’t suddenly decide to trust one another. Since neither is prepared to lower its guard, there are still many things that could go wrong. A major flare-up over US sanctions on Iran, Chinese support for Russia, or a miscalculation over Taiwan could raise pressure in both capitals for a more confrontational approach. But as long as the US and China continue to read the balance of leverage correctly and avoid domestic pressure for escalation, the occasional diplomatic, trade, or security skirmish can remain contained in the interest of stability.
There is a dangerous wildcard, though. AI and the lack of international governance over its development and use are already creating risks, within both the US and China and around the world. And given that the US and China now lead in AI innovation by some distance, competition between the two in this sector poses the biggest threat to their currently constructive relations.
This race is now marked by both fast-advancing Chinese capabilities and the sudden rise of new national-security risks for both countries. China’s decision to focus on inexpensive, open-source models will make AI at scale accessible to consumers and industrial users in the West and across the Global South. But the US intelligence and defense community will not welcome such competitive pressure (and data collection). Moreover, Chinese models bring some of the same cyber and biotech risks as their American counterparts; but unlike Anthropic or OpenAI, Chinese companies won’t respond to US government concerns and demands.
That is why the US will respond in the coming months with significant restrictions on Chinese access to US models and limits on US access to Chinese open-weight models. Trump will also try to use US political leverage to pressure other countries to adopt American, rather than Chinese, AI. If the US and China fail to manage this struggle, we can expect a sharp acceleration in the decoupling of advanced technologies between the two countries.
We are already seeing this trend in other areas. In July, the US banned the import of foreign-made humanoid robots—a move obviously aimed at China, which has a substantial technological and pricing lead over its American competitors. Trump has also now imposed up to 100% tariffs on drones, targeting Chinese-made models, in particular, on national-security grounds.
In short, the Sino-American AI arms race is already well underway. The question now is whether Trump and Xi can use their White House summit this month to lay the groundwork for an ongoing AI dialogue designed to avoid a future of accelerating, opaque, and potentially zero-sum conflict.
If they fail, the two sides may feel that confrontation has become inevitable, and not just regarding AI. It is not hard to imagine a renewed trade conflict and heightened security tensions quickly bringing down the guardrails the two sides have built over the past 16 months. The AI race makes such an outcome all the more difficult to avoid.