Invest1 publisher3 min readPublished
Trump and Xi refuse the AI speed limit ten days before they meet
OpenAI and Anthropic want Washington and Beijing to negotiate a pace for frontier AI development. Trump's answer arrived on Truth Social on Sept. 14, and China's Foreign Ministry called the U.S. debate fear-mongering the same day.
The Investor · Invest desk
What happened
- The CEOs of OpenAI and Anthropic renewed their call for Washington to coordinate with Beijing on slowing AI, after a former Anthropic researcher's viral X post claimed the technology could kill all of humanity by the end of the decade.
- Amodei's Sept. 12 letter asks the U.S. and China to agree to a "speed limit" for AI development, and in a CBS Sunday Morning interview he likened the rivalry to the Cold War nuclear race.
- Trump answered on Truth Social on Sept. 14 that the only guardrails the U.S. needs is a strong and smart president, and added, "Whoever wins AI, wins!"
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- constraint With no government-to-government cap in prospect, a lab that slows its own training runs hands share to one that does not, and the speed limit has no enforcement path in either capital.
- exposure Anthropic's safety argument is now heard in Beijing as commercial protection, so its later claims about model risk arrive there discounted by the chip-export term attached to them.
- precedent The realistic output of Sept. 24 is narrow: language against weaponizing AI and principles on non-state-actor misuse. That language becomes the template any later talks build from.
- contradiction Fortune's own sources split on Xi's appetite, with Triolo describing a leader who wants frontier safety talks and Babic describing framing too conniving to shape Xi's agenda at all.
A speed limit on frontier training is a cap on capital spending. A cap that one lab honors while its competitors keep spending costs that lab share. So Sam Altman and Dario Amodei asked the two governments that could make it apply to everyone [2]. The answer came in two days [20]. Trump wrote that "the only controls or 'guardrails' the U.S. needs is a STRONG AND SMART (High IQ!) PRESIDENT" [4]. The same day, China's Foreign Ministry spokesperson Guo Jiakun called the American discussion "fear-mongering" that "will only hamper efforts toward sound global AI governance, which serves no one's interest" [6].
Amodei's Sept. 12 letter bundles the pace request with an export position: a floor under which neither country uses AI to develop biological weapons, plus his restated view that the U.S. should not sell advanced chips to China [1][10]. In China the slowdown campaign has been received as an attempt by Washington to hold its lead, and the chip export limits already pursue that lead [21]. Brosi Babic, a professor at the University of Hong Kong, calls the letter "self-serving editorializing" that is "conveniently coming at a time when the gap between Chinese and frontier US models is shrinking, as an attempt to hang on to a vanishing market lead" [11]. A post by Shengyu Liu, an engineer at DeepSeek, is gaining traction for comparing Anthropic reaching "advanced artificial intelligence" to "Hitler obtaining atomic-bomb technology" [18].
George Chen, partner and chair of digital practice at The Asia Group, described the incentive on the Chinese side. "Xi is unhappy with recent U.S. moves to contain Chinese advances in AI, robotics, and drones," Chen said [13], and "China does not need U.S. permission to accelerate or decelerate its AI investments; Xi will pursue his own agenda" [14].
The American rejection was delivered by phone. Trump called Jensen Huang while the Nvidia chief executive was speaking on stage. Huang put him on speaker in front of the crowd, and Trump said, "Whatever Dario said this weekend won't stop our progress" [7]. He also called the backlash to data centers and fears that AI will "take over" a "hoax" [8].
Altman's approach was prestige. He told Fortune on Friday that he believed Trump and Xi could win the Nobel Peace Prize if the two leaders struck a deal on AI safety [9]. Neither leader forfeited anything the sources can price by declining.
Which leaves the pace between now and the meeting to the labs' own capital plans, since the only two parties who could set it jointly have both said no [3]. Paul Triolo of DGA, who says any agreement is "a long way" off, also says Xi wants a "serious dialogue on frontier AI model safety" [15][16]. A narrow start can still bind: joint language on not weaponizing AI, or the biological weapons floor Amodei calls the bare minimum, would mean intergovernmental coordination is live and this argument is wrong [17][10]. Or Beijing sets a pace of its own, on its own terms; Chen's point about Xi's agenda leaves room for that [14].
What to watch
- Whether the non-state-actor misuse principles Chen describes turn up in any text published after the Sept. 24 meeting.
- Whether Anthropic keeps the no-advanced-chips-to-China term that Babic says makes the letter unpersuasive in Beijing.
- Whether either government names a baseline for which frontier models get tested, the gap Triolo identifies.