
Bessent says Trump and Xi agreed humans must stay in major military AI decisions
Bessent says Trump and Xi agreed in Shenzhen that humans must remain in major military decisions involving advanced AI. Chinese readouts affirm human control in more general terms. No treaty text or enforcement mechanism has been published.
Documented: On Nov. 20, 2026, U.S. Treasury Secretary Scott Bessent told CNBC that President Donald Trump and Chinese President Xi Jinping, meeting the day before in Shenzhen on the sidelines of APEC, agreed that humans should always be involved in military decisions made with what the Trump administration calls “super intelligence,” and that the technology should not be allowed to act autonomously in major military actions. Reuters, Bloomberg, and the South China Morning Post reported the remarks. Bessent described the understanding as one of the main results of the Asia trip and said it was separate from the two countries’ competition for AI leadership. He also said U.S. labs should stay at the frontier as a defense against what he called dangerous Chinese models used in cyberattacks. No published treaty text, definitions of “major military actions,” or enforcement mechanism accompanied the interview.
Documented, narrower on the Chinese side: China’s Foreign Ministry readout of the Shenzhen meeting, as reported by China Daily and ThinkChina and summarized by China-tech analyst Kyle Chan, said the two sides should strengthen communication to guard against AI safety risks and ensure AI remains under human control. That language is general. It does not, in the public Chinese account, spell out a ban on autonomous launch of major military operations.
Documented, also from the U.S. side: Trump told reporters the two leaders would work on “joint guardrails,” without specifying military systems, technical standards, or a timeline. China Daily reported that he did not detail the proposed arrangement. Bessent separately described robust talks on trade and on containing the growth of nuclear weapons. A trilateral meeting with Russian President Vladimir Putin produced a restatement that a nuclear war cannot be won and must never be fought, according to wire accounts of the joint declaration — a nuclear line, not an AI protocol.
Claimed, not independently verified as a binding pact: That the leaders “agreed” a specific rule that SI “should not be allowed to be autonomous for major military actions.” That formulation is Bessent’s account of the talks. Trump’s public language was about joint guardrails and, in other remarks reported around the trip, moderate safety regulation that would not destroy commercial value. Chinese public language stayed at human control and risk coordination. Secondary write-ups that checked CNBC, Reuters, and related coverage found no cited implementation timeline.
Speculated, and not supported by the public record: That this is a formalized rules-of-engagement regime for the next century of warfare, or a completed superpower treaty on existential risk. What is on the record is a high-level political understanding, described mainly by one U.S. cabinet official, sitting next to continued chip controls, an explicit U.S. determination to stay ahead in the technology race, and no published verification regime. The interesting fact is the split: both governments are willing to say, in different registers, that major uses of advanced AI should remain under human control — while neither has published the operational meaning of that sentence.
Militaries, AI labs, and anyone living under U.S. or Chinese power are affected by whether human sign-off on major uses of military AI is a real constraint or only a talking point. What is documented so far is a political statement from both sides, not a published rule with enforcement.
Sources (6)
- [1]Trump, Xi agreed that military AI decisions should involve humans, Bessent says(https://www.reuters.com/world/china/trump-xi-agreed-that-military-ai-decisions-should-involve-humans-bessent-says-2026-11-20/)
- [2]Trump, Xi Agreed on AI Curb in Military Action, Bessent Says(https://www.bloomberg.com/news/articles/2026-11-20/trump-xi-agreed-on-ai-limits-in-military-action-bessent-says)
- [3]Treasury Sec. Bessent: Humans should be involved in any military decision(https://www.cnbc.com/video/2026/11/20/treasury-sec-bessent-humans-should-be-involved-in-any-military-decision.html)
- [4]Xi, Trump agree military AI decisions must involve humans, Bessent says(https://www.scmp.com/news/china/diplomacy/article/3375321/xi-trump-agree-military-ai-decisions-must-involve-humans-bessent-says)
- [5]Trump reviews Asia trip, praises China ties and discusses joint AI safeguards(https://global.chinadaily.com.cn/a/202611/21/WS6b0138dbe4b03ebf8519ddd9.html)
- [6]Bessent says Trump and Xi agreed humans must stay involved in military AI decisions(https://casrai.org/news/trump-xi-human-oversight-military-ai-bessent-2026)