China’s account of a White House meeting says Xi Jinping called for AI to remain under human control, while Donald Trump separately signalled resistance to additional rules.

China’s Ministry of Foreign Affairs published its account of the presidents’ 24 September meeting the next day. It says Xi proposed continued US-China dialogue on AI, exchanges about risks and benefits, joint action against misuse, and a people-centred approach. Its clearest line was that AI must remain under human control.

The same account says Trump supported maintaining dialogue and strengthening AI cooperation. It does not say he accepted Xi’s language about human control or a specific regulatory programme. The Washington Post separately reported that Trump wrote on Truth Social that AI would be discussed but that he wanted to leave it “exactly where it is.” This review could not verify that social-media post from a primary White House transcript.

Why it matters

The exchange puts two different policy instincts beside each other. Xi’s statement frames advanced AI as a shared-risk problem requiring bilateral discussion and an explicit human-control principle. Trump’s reported post frames new constraints as unnecessary. That gap matters because the United States and China host many of the companies, researchers and computing resources that would have to implement any practical agreement.

Dialogue is not a control mechanism by itself. Terms such as “human control,” “misuse” and “abuse” need operational definitions. A policy could require human authorization before a weapon is fired, for example, while still allowing software to select targets or compress the time available for review. It could also mean keeping people formally in a workflow without giving them the information or time needed to override a model.

The official Chinese account supplies no proposed test, monitoring system or enforcement route. It does not identify which classes of model would be covered, how incidents would be disclosed, or what either government would do if a company or agency crossed a boundary. It records a diplomatic position, not a negotiated standard.

Cooperation and competition remain intertwined

The Chinese readout says both countries have strong AI capabilities and may compete, but can cooperate more. That formulation leaves room for technical exchanges on evaluation, incident reporting or authentication while strategic competition continues. Shared work could be narrow: common terminology, notification of serious failures, or communication channels during an AI-related security incident.

The obstacle is verification. Governments may be unwilling to reveal model capabilities, military uses, training infrastructure or incidents that expose intelligence methods. Private laboratories also control much of the relevant technical evidence. An agreement without access to logs, evaluation methods and named responsible institutions could remain a statement of intent.

Trump’s reported resistance to new rules does not necessarily rule out all cooperation. The Chinese account attributes to him support for AI dialogue. Technical coordination can occur through procurement terms, voluntary laboratory protocols or security communications rather than a broad new law. But the difference between cooperation and enforceable restraint should remain visible.

There is also a source asymmetry. The detailed meeting account comes from China’s foreign ministry and presents the Chinese government’s selection and wording. The reported Truth Social post comes through a newspaper. Neither is a jointly agreed transcript. Claims about what the leaders accepted should therefore be narrower than claims about what each side says was discussed.

The next meaningful evidence would be a bilateral statement naming a workstream and deliverables. A shared definition of human control or a timetable for expert meetings would turn diplomatic language into something testable. Until then, the meeting shows that AI risk reached the presidential agenda, not that the two governments settled how to manage it.

Verification

  1. VERIFIED — Xi and Trump met at the White House on 24 September, according to China’s official account. Primary source: https://www.fmprc.gov.cn/eng/xw/zyxw/202609/t20260925_12031181.html
  2. VERIFIED AS CHINESE GOVERNMENT STATEMENT — The readout says Xi called for AI dialogue, action against misuse and human control. Primary source: the Chinese Ministry of Foreign Affairs page above.
  3. VERIFIED AS CHINESE GOVERNMENT ATTRIBUTION — The readout attributes support for continued AI dialogue and cooperation to Trump. Primary source: the same page.
  4. UNVERIFIED FROM A PRIMARY TRANSCRIPT — Trump’s reported Truth Social wording was not available in a White House transcript reviewed here. Via: https://www.washingtonpost.com/technology/2026/09/25/trump-rejects-demands-ai-rules-while-xi-calls-human-control/
  5. ANALYSIS — Possible technical forms of cooperation and their verification problems are editorial analysis.

Glossary candidates

  • Human control: Meaningful human authority over whether and how an automated system acts.
  • Incident reporting: A process for disclosing serious failures, misuse or security events.
  • Readout: An official government’s summary of a diplomatic meeting.

Cold-reader sentence: Xi proposed human control and bilateral AI dialogue, but no joint mechanism or enforceable standard emerged.