US officials pitch an AI-incident hotline between Bessent and He Lifeng at the Trump-Xi summit

The board-deck version of the Trump-Xi summit ^[1] is a US-China brake on AI. So far the record supports a phone line. In the Vox account, Washington and Beijing "could" set up the hotline American officials outlined ^[2], and the author writes that the summit "may be a step in the right direction" ^[14]. We do not know yet whether Beijing will agree to it.

The proposal puts incident response at the top of each government, with the US Treasury on one end and a Chinese vice premier on the other ^[2]. The government bodies that actually examine models are the national AI safety and security institutes. These government-funded groups test companies' latest products before release ^[5].

The hotline would sit on top of the four mechanisms already built ^[6]. Under the G7-led reporting mechanism, AI companies share how they are building their latest models, and the mechanism also sets standards for reducing catastrophic risks. Separately, an international scientific report updates governments each year on risks from the most advanced systems ^[5]. All of this came out of four years of talks in Congress, the White House, the UN and the G7, and the work picked up after ChatGPT arrived in 2022 ^[12]. The author, a technology journalist now at a think tank ^[13], wrote that "we already have a quasi-planetary shield," but that "it's not really up and running yet" and that it is "not clear whether it will actually work" ^[7].

The author wants those pieces turned into "a functioning, but crude, first-responder system when things suddenly go wrong" before the end of 2026 ^[8]. At present there is no way to respond globally and in real time when an AI crisis cuts across countries already skeptical of each other ^[9].

Some of the urgency comes from the companies themselves. "We could lose control of the future to AI," OpenAI's Sam Altman told the UN Security Council ^[4]. He spoke on the day heads of leading AI companies asked the UN to police the technology ^[3]. The Vox author is the skeptic in this account. "I still think many of the concerns outlined by the AI leaders themselves are massively self-serving," the author wrote, "especially when their companies are soon to go public" ^[10]. The same author says the growing list of AI scandals is real, and that it shows the current way of regulating the technology is not working ^[11].

For an operator, the useful split is between what is on the table this week and what gets built after it. This week's item is a line between two officials. The Vox piece asks directly what kind of regime works "when there is no trust between major governments" ^[15]. A phone line is a modest answer to that question. The rules that already reach a company come from the four existing pieces ^[6]. The author's timetable for turning them into a crisis-response system runs to the end of 2026 ^[8]. I'd expect those rules to keep arriving in pieces from different bodies, as they have for four years ^[12].