US Treasury Secretary Scott Bessent said Presidents Donald Trump and Xi Jinping agreed that humans must always be involved in military decisions using artificial intelligence. Bessent called the understanding one of the major achievements of Trump’s Asia trip, reached during the leaders’ meeting in Shenzhen on 19 November.
What Was Agreed
Bessent told CNBC that the two leaders agreed humans should always be involved in any military decisions that so-called superintelligence makes. The technology should not be allowed to act on its own in major military actions, he said, using the administration’s preferred term for advanced AI.
“There is nothing to do with competition when it says that both sides agree that no major military operation should be launched without a human being involved,” Bessent said. He added that the guardrail would stand apart from the two countries’ race for AI supremacy.
The Shenzhen Meeting
The agreement came during talks on the sidelines of the Asia-Pacific Economic Cooperation summit in Shenzhen, the southern Chinese tech hub. It was the third in-person meeting between Trump and Xi this year, after Trump’s state visit to Beijing in May and Xi’s trip to Washington in September.
Chinese state news agency Xinhua reported that Xi told Trump the two countries should work together to prevent security risks linked to AI and keep the technology under human control. The leaders are expected to meet again next month at the Group of 20 summit in Miami.
Chip Curbs Stay in Place
The understanding does not soften Washington’s technology restrictions. Bessent said Nvidia’s H200 remained the most advanced chip China could buy under US licensing, and that Chinese orders for it had been scarce as Beijing builds its own domestic chip stack.
The Commerce Department moved on Monday to tighten China’s access to advanced AI chips further, including chips used in overseas data centres where Chinese firms can rent computing power. Bessent said the proper restrictions were now in place and remote access had been curtailed.
Our Read
This is a norm, not a treaty. No joint text was published, no enforcement mechanism was described, and no definition of a major military operation was given. Its weight rests on two leaders saying the same thing in public, through Bessent and through Xinhua, rather than on anything signed.
Even so, the signal matters. Military AI is the sharpest edge of the US–China technology race, and an informal shared rule against fully autonomous major operations gives both sides a reference point in a crisis. The test will be whether the understanding survives the next round of chip restrictions.
The chip fight and the safety talk now run on parallel tracks. Washington is tightening access to advanced hardware while agreeing, in principle, that neither side should hand major military decisions to machines. That split — compete on capability, cooperate on control — is the deal’s real shape.
Why It Matters
A human-in-the-loop rule for military AI, agreed by the world’s two largest AI powers, could slow the slide toward autonomous weapons even without a treaty. For the industry, it frames the safety debate concretely: control of military systems is where the two sides found common ground.
Background
The agreement follows months of tightening US controls on advanced chips and AI models bound for China, alongside parallel talks on AI safety. Earlier this week, reports said the Trump team was weighing rules forcing firms to disclose their use of Chinese AI systems.
Source: Reuters report and SCMP coverage.


0 Comments