Washington drafts a letter turning 35 AI signatures into a forced choice

Reuters' Michael Martina reported on 14 August that the United States has drafted a letter to the 35 signatories of its June "AI Opportunity Statement" telling them to pick a side on artificial intelligence, citing a US official and an internal document ^[1][2]. The statement those countries signed commits them to nothing more than a "common purpose" and a "shared vision" with Washington, by the State Department's own description, and the framework is non-binding ^[9][10]; the letter is an attempt to turn that into a commitment ^[3].

The language is unusually direct for a diplomatic text. "To be part of everything is to be part of nothing," the draft says, adding that signing the Pax Silica Declaration is "not merely a membership subscription, but a commitment" ^[3]. China is never named. The draft refers instead to "duplicative initiatives whose expectations conflict with our own" ^[17]. The US official Reuters spoke to, on condition of anonymity, was blunter: the letter was drafted to make clear "you can't have it both ways", and the official said it is difficult to see how a country can credibly position itself as a trusted partner in one technology ecosystem while joining an initiative designed by China to advance a competing vision ^[18].

What makes this an operating problem rather than a diplomatic one is Pax Silica's scope. Washington launched it last year to secure supply chains across three things at once: AI models, semiconductors and critical minerals ^[6]. Members get access to shared investment opportunities in AI projects, and the framework pushes them towards joint projects and coordinated export controls ^[8]. About two dozen countries have joined, with Reuters naming Japan, Australia, South Korea and Kazakhstan ^[7]. That leaves roughly eleven signatories of the looser statement who are now being asked to upgrade ^[1].

Kazakhstan is the reason. It joined Pax Silica in June, an accession the State Department highlighted through Under Secretary Jacob Helberg as the first in Central Asia, and it holds significant reserves of the minerals advanced chips depend on ^[11][12]. It then also joined China's framework, making it the only country known to have signed both, which Reuters reports set off alarm bells in Washington ^[13]. Beijing announced that rival body, the World Artificial Intelligence Cooperation Organization, at a Shanghai conference in July, with Xi Jinping pitching China's open-weight models as the alternative to American influence ^[15]. China has already used its near-monopolies in critical minerals as a retaliatory weapon during the tariff war ^[14].

The timing tracks a shift in the technology itself. Chinese open-weight models have closed much of the gap on the proprietary systems from OpenAI and Anthropic ^[24], and Hugging Face's chief executive told CNBC this month that China leads the open-model race with 41% of downloads ^[19]. Nvidia and several other large firms broke with OpenAI and Anthropic over open weights this year ^[20]. Enforcement is tightening in parallel: the Commerce Department is examining how China obtains remote access to Nvidia chips it cannot buy, and the Treasury Secretary has threatened to sanction Chinese AI firms outright ^[21][22]. Beijing's own openness is conditional too, with Reuters reporting in July that it is weighing restrictions on overseas access to its leading models ^[16].

Three things to watch. Whether the letter is sent at all, and in what form: the draft is undated, Reuters could not establish timing or whether the text changes, and the State Department declined to comment on "purportedly leaked internal documents" ^[4][5]. Whether Kazakhstan is asked to withdraw from one framework, which would set the precedent for every hedging supplier ^[13]. And whether any European country turns up on the membership list, given that Reuters named none, though it did not publish the full roster ^[23].