Xi Jinping stood before an artificial intelligence conference in Shanghai on Friday and delivered what, on its surface, sounded like a manifesto for the open internet: share the models, democratize the technology, stop weaponizing national security to hoard semiconductors and code. It was a fluent performance — and a strategically timed one.

The timing is not incidental. Washington has spent the better part of two years tightening export controls on advanced AI chips, specifically the high-bandwidth memory and cutting-edge GPU architectures that power frontier model training. The U.S. Commerce Department has expanded its Entity List restrictions multiple times since 2022, explicitly naming Chinese labs and procurement networks. Xi's speech was, at minimum, a direct public rebuttal to that policy architecture — framed not as national grievance but as a universal principle of fairness.

That framing deserves scrutiny. China's domestic AI ecosystem operates under a regulatory regime that requires large language models to be registered with the Cyberspace Administration of China before public release, that mandates content controls aligned with 'core socialist values,' and that gives state security organs broad access to data held by domestic technology companies under the 2017 National Intelligence Law. Openness, within China, has a ceiling — and that ceiling is the Party.

None of that makes the open-source argument wrong on its merits. The global AI research community has made a credible case, independent of Beijing, that open weights and open architectures accelerate safety research, reduce monopoly concentration, and give smaller nations a seat at the table. Xi is surfing a genuine intellectual current — he did not invent it. But surfing it at a state conference, with Chinese diplomatic influence operations running in the background, is a different act than publishing model weights on GitHub.

What China has actually accomplished in the open-source space is real and should not be minimized. DeepSeek's R1 model, released earlier this year, demonstrated benchmark performance that rivaled closed frontier models at a fraction of the reported training cost, and its weights were made publicly available. That release genuinely shook assumptions in the U.S. technology industry about the cost structure of frontier AI. Whether it was a product of brilliant engineering, state subsidy, or some combination is still being parsed — but the capability was not fabricated.

The chip controls are the subtext of everything Xi said on Friday. The U.S. argument is that advanced AI compute has direct military application — autonomous targeting, signals intelligence, large-scale surveillance infrastructure — and that allowing adversary states to acquire it freely is a national security risk of the first order. China's counterargument, now voiced by its head of state, is that securitizing AI development is itself a destabilizing act that entrenches existing power and leaves the developing world dependent on American corporate infrastructure. Both positions contain real arguments. Neither is purely principled.

What Xi's speech does, strategically, is position China for the Global South. Countries in Southeast Asia, Africa, and Latin America that are navigating their own AI procurement decisions now hear Beijing offering access and openness while Washington offers restrictions and licensing conditions. That is a competition for influence dressed in the language of technology policy, and it is one the U.S. foreign policy establishment has been slow to engage on its own terms.

The honest read of Friday's address is this: Xi Jinping made a geopolitically useful argument that happens to overlap with a philosophically defensible position. The overlap is not a coincidence, but it also does not automatically make the argument false. The question worth asking — the one the conference hall in Shanghai was not designed to answer — is whether any government that surveils its own citizens at scale, restricts what its own researchers can say publicly, and treats data as a state asset has standing to lead a global conversation about AI openness. That question will not resolve itself on its own.