In a long essay published on September 12, Anthropic CEO Dario Amodei argued that AI companies and governments should jointly slow the pace of frontier model capability gains, leaving more room for safety and alignment work. Sam Altman and Elon Musk both said they agreed. The appeal was aimed at the industry itself, but within three days, the two loudest responses it drew came from Washington and Beijing — and they pointed in the same direction.
A Response From the Fairways of Ireland
On September 13, Trump addressed the matter with reporters at his Doonbeg golf resort in Ireland, pushing the warning aside. His exact words:
We're leading China in AI, we're the most sophisticated country in the world, and frankly, I want to keep it that way because whoever wins AI, wins.
He also attributed outside concern over AI risk to "very negative forces," and suggested some parties were using AI and data centers as a political target. Based on public reporting, the White House has not issued any document or executive order on the matter — this was an off-the-cuff remark to reporters, not a policy action.
Beijing's Word for It: 'Fear-Mongering'
On September 14, Chinese foreign ministry spokesperson Guo Jiakun addressed the same issue at a regular press briefing:
Fear-mongering, confrontation, competition will just disrupt [the] process of global AI governance.
The Global Times was more direct, saying the appeal was "packed with containment clauses aimed at China" and calling it a Cold War playbook for AI. Some reports say a Chinese security official expressed a similar view, framing the proposal as an attempt to hold back China's development — though that account currently exists only in secondhand reporting, and no original text has been made public.
The common thread in both statements is clear: neither side treats slowing down as something up for discussion, and each is using the other to explain why it can't afford to. Washington's reasoning is that it can't let China catch up; Beijing's is that the proposal was aimed at China from the start. Each argument doubles as evidence for the other.
What's Left of the Original Proposal
Amodei's plan does include actionable steps. Anthropic has already put its commitments in writing: office space and building access for third-party evaluation teams, access to its internal risk team, and publication rights free of editorial control. OpenAI has said it will follow through on the first step, but no terms or timeline have been released. The remaining two steps of the three-step plan have seen no movement so far.
What the two companies have actually offered isn't equivalent: Anthropic has handed over specific, verifiable arrangements, while OpenAI remains at the level of a statement of support. Behind the same appeal, there's a real gap between how many people signed on and how much has actually been put on paper.
This path operates on a different level from the political statements. The corporate track runs on voluntary commitments plus third-party verification and doesn't require legislation; the political track determines whether any top-down force will compel it. The two statements on September 13 and 14 effectively closed off half of that external-pressure option from both directions at once.
What It Means From Here
The fallout from this exchange will land quickly on the industry side. Washington's unwillingness to tighten rules means frontier labs' release schedules won't slow down under regulatory pressure; Beijing's framing of the appeal as containment largely rules out any US-China dialogue on a "joint slowdown." With a summit between the two countries' leaders approaching, AI will likely show up on the agenda as a competitive issue rather than a governance one.
For developers and enterprise users, the practical takeaway is that nothing changes pace-wise: model iteration, price cuts, and capability jumps will continue on each company's own schedule, with no external brake being installed. The signal worth watching instead is different — which company, if any, pushes back an already-scheduled release, by how long, and with what explanation. That's the only thing that would actually confirm the appeal has teeth, and so far no company has done it publicly.
Sources: NPR, CocoLoop, CNBC, Global Times; quotes from Trump and the Chinese foreign ministry spokesperson were checked against the original English wording in each outlet's reporting.