Invest4 publishers2 min readPublished Updated
Xi skips the UN General Assembly to negotiate AI with Trump in Washington
Twenty-two countries adopted a UN declaration on human oversight of AI systems this week. The United States and China spent Sunday discussing a bilateral risk-notification channel of their own.
The Investor · Invest desk
What happened
- A.I. dominated a set of U.N. meetings in New York on Monday centred on international digital cooperation, covering the technology's risks, economic dislocation and possible international governance.
- Xi Jinping is expected to skip the General Assembly and convene instead in Washington with President Trump during a state visit this week.
- The United States and China discussed on Sunday creating a notification system that would let each country alert the other about A.I.-driven national security risks.
- Leaders of 22 countries adopted a declaration on Monday calling to strengthen human direction, oversight and control of A.I. systems. The United States and China did not join.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint AI firms everywhere else take their binding rules secondhand: advanced chip supply is licensed in Washington, and intellectual property exposure is defined by the same two governments.
- decision Anyone planning compliance around an international AI standard now has to track a bilateral channel that so far exists only as a name, and pick a date to check again.
- contradiction Guterres says voluntary pauses fail unless everyone agrees, and the two governments whose agreement he means were building a separate channel the day before the declaration was adopted.
- precedent Beijing's new Shanghai organisation makes the contest one over which multilateral body hosts AI rules, so the next fight is about venue and membership.
Monday's declaration asks 22 governments to strengthen "human direction, oversight and control" of AI systems [6]. The instruments that already reach AI firms sit elsewhere: the American restrictions on advanced computer chips that Beijing keeps objecting to [9], and the American accusations that Chinese AI firms stole intellectual property from US labs [8]. Both are administered by Washington and Beijing [7].
The bilateral channel got a name on Sunday. Bessent said the notification mechanism would be called "the U.S.-China A.I. dialogue" [5]. The New York Times report does not say how the mechanism would be triggered, verified or started [22], and the administration proposing it is led by a president who has repeatedly dismissed AI safety concerns as a hoax [2]. For a firm budgeting compliance work, the dialogue is so far something to monitor.
The UN has been on the file longer. It formed an advisory body on AI in 2023 [13], and the declaration arrived about three years later [21]. Guterres last week placed AI alongside climate change and deepening global inequality as one of three existential risks facing humanity [10]. "The world cannot afford a race to the bottom on A.I. safety," he said [11]. He also said voluntary pauses on the technology would not work if not everyone agreed to them, and that means the United States and China [12].
Kyle Chan, a foreign policy fellow at the Brookings Institution, said the UN "has been sidelined in the A.I. conversation" [14]. He called AI "a two-player game" and added: "It's kind of like nuclear weapons: If you do not have them or are not on the verge of having them, it's hard to be relevant in shaping the discussion" [15]. Foreign policy experts told the Times the rest of the world has largely been pushed to the periphery [18].
Beijing does not follow that logic all the way. It launched the World Artificial Intelligence Cooperation Organization this summer, headquartered in Shanghai, in a bid to assert global dominance on international AI discussions [16]. Washington, the Times reported, has shown little interest in building international consensus [17].
Maybe the dialogue acquires published terms, and the New York process has something concrete to attach itself to. Or it stays a name, and the only enforceable constraints on AI firms remain chip licences and intellectual property claims [9][8]. Or the Shanghai organisation collects members and the UN loses the file to a rival forum. In my view the second is most likely. If the language in Monday's declaration turns up as a condition inside a national export licence, that view is wrong.
What to watch
- Whether either government discloses the first invocation of the U.S.-China A.I. dialogue.
- Whether the Shanghai-headquartered World Artificial Intelligence Cooperation Organization announces members beyond China.
- Whether the United States or China later joins the 22-country declaration on human oversight of AI.