While American executives were signing their letter, the word at the heart of it was being claimed just as fervently in Shanghai. A week earlier, at China's World AI Conference, President Xi Jinping made his first appearance at the event and delivered a keynote that borrowed the very language US industry now treats as its own. He called for "a symphony of global cooperation" instead of "a solo performance by any single country," urged the world to "jointly oppose overstretching the national security concept" in AI, a pointed jab at US export controls, and pledged 5,000 AI training slots for developing nations.
The rhetoric arrived with institutions attached. The day before Xi spoke, twenty-nine countries, including Russia, Pakistan, Indonesia, Brazil and Kazakhstan, signed an agreement in Shanghai establishing the World Artificial Intelligence Cooperation Organization, a body headquartered in the city and explicitly meant to rival US-led AI governance. The UN Secretary-General was in attendance. Days later, twenty-one APEC economies, the United States and China among them, issued a "Chengdu statement" endorsing open-source development, the first time such cooperation has been affirmed at ministerial level.
But read the fine print and the openness on offer is not the libertarian kind. The Chengdu text wrapped its support in language about "security, data protection and intellectual property rights," and specifically backed models "that employ strong security assurance through development and deployment." That is open source drifting toward state supervision. As one analyst put it, the debate is moving past open versus closed toward a harder question: who can build an open ecosystem trusted enough for governments and enterprises to actually deploy.
For Washington, the timing is awkward. The American industry's letter frames openness as a distinctly national inheritance, the legacy of 1980s open-source software. Yet China now leads the open-weight field, its Kimi K3 and GLM models sitting near the frontier and free to download, and Xi is now occupying the same pro-openness, anti-restriction ground from a rival capital. It is a rhetorical space the US coalition assumed was theirs, and it has been quietly annexed.
The response inside the administration is anything but unified. AI adviser David Sacks has warned against what he calls the "Kimi Panic," arguing that America "ties itself in knots" with rules while competitors advance, and that leading labs simply "want the government to eliminate their open-source competition." Others want precisely the opposite: tighter controls on Chinese models, which by one measure already account for roughly 60 percent of the token usage US developers run through the OpenRouter platform. Sanctions over alleged distillation are on the table; so is a planned US-China AI dialogue in September that the feud could derail.
What is emerging is a contest not only over who builds the best models but over who owns the idea of openness itself. Beijing frames it as sovereignty and access for the global south. Washington frames it as competition and American leadership. Both are describing the same technology with the same word. A movement that began as a rebellion against corporate control of code is now a banner carried, on opposite sides of a widening divide, by two of the most powerful governments on earth.