A view from DC: China promotes AI sovereignty and open models

In Shanghai this week, China hosted what was billed as the 2026 World AI Conference and High-Level Meeting on Global AI Governance. In a major speech, Chinese President Xi Jinping outlined the country’s vision for the AI era, positioning his country as a collaborator in the global community, even as it seeks to lead the way.

In a pointed reference to the U.S. approach, Xi first recited a Chines proverb, "A single string cannot make music, and a single tree does not make a forest," then added, "AI development should not be a solo performance by a single country, but a symphony of international cooperation."

A primary driver behind the 2026 conference appears to be the creation of the World Artificial Intelligence Cooperation Organization. Headquartered in Shanghai, it is the world's first intergovernmental international organization focused on AI. Twenty-nine countries signed the founding agreement during the conference to promote open-source collaboration, bridge digital divides and ensure AI is safe and accessible globally.

Though seemingly focused higher up the tech stack, the WAICO appears to be a response to the U.S.-led Pax Silica initiative to secure semiconductor and critical mineral supply chains, which features its own set of 24 signatory countries. Pax Silica has already established initiatives like the "Pax Pass" for fast-tracking related shipments through the Panama Canal and a major manufacturing hub under construction in the Philippines. 

Separately, the U.S. Department of Commerce has been focused on launching its AI Exports Program, announcing that 78 companies have submitted applications under the program. But Politico reports the program's results are not meeting expectations so far, quoting unnamed industry executives as indicating, "Many of the country's biggest AI companies remain unconvinced the government's financing, advocacy and licensing incentives will meaningfully boost foreign sales."

He who controls the standards controls the world

In his speech, Xi directly compared the rise of AI with other major technological advances. "In the course of history, the invention of the steam engine heralded the industrial civilization, the widespread access to electricity brightened up modern society, and the birth of the Internet brought the entire world together. Each of these technological revolutions has profoundly reshaped our way of work and life, and enabled a giant leap in economic and social development."

What he did not say explicitly was that each of these technological revolutions was also marked by battles over standard setting and adoption. Whether you look at the race toward regulatory capture for interconnection rules in the build-out of railroads or the messy and misinformation-fueled war between the proponents of AC and DC currents, the long-term winners are always those who pioneer the winning standards in each technological revolution. 

China understands this maxim is just as true for countries in our interconnected geopolitical era as it was for individual companies during the gilded age. And so, China is staking its claim over open models, framing them as a global public good.

From open door to open weights

The historical irony here is that the open-source philosophy was initially incubated within Western research institutions as an ideological commitment to algorithmic transparency, peer review and distributed innovation. Over recent years, however, intense commercial consolidation and national security anxieties have prompted a contraction in the U.S. embrace of openness. As one example, you see this in Meta's cautious recalibration of its previously aggressive open-model strategy.

In some ways, the philosophy underlying open models is a direct intellectual descendant of the open-source software movement. But it has major distinctions, especially because open models do not reveal the underlying inputs and training logic, which is arguably the true source code of advanced AI systems.

What proponents of openness do argue, however, is that open models allow for local autonomy without centralized control, better verification for research purposes, access to expensive infrastructure that would otherwise be cost-prohibitive for most to develop, and security through mass scrutiny of alignment failures.

Each of these arguments map nicely into the geopolitical context and resonate with under-resourced countries attempting to adapt quickly to the AI era. Beijing offers developing nations the promise of absolute control over their own digital infrastructure. Of course, from a skeptical perspective, adopting these models still aligns the client economy with the underlying architecture and hardware requirements dictated by China. Xi frames this strategy as protecting against "new historical injustices" in AI, but it is unclear how much autonomy will really be delivered.

China is also capitalizing on the global response to recent U.S. steps that demonstrated why reliance on a centralized developer infrastructure can put economies at a disadvantage. As Politico reported about the supposedly underwhelming response to the Commerce Department full-stack initiative, executives are "unnerved by the administration’s abrupt, since-reversed order forcing leading AI company Anthropic to cut off foreign access to its latest models — a move they warn may undercut the very pitch the program is making abroad."

There is also a practical reason for openness. Beijing can't afford the compute power necessary to centralize generative AI infrastructure. Distributing open weights offloads compute to the edges, placing those costs on the client countries. But this strategy is more than words. Just this week, Chinese startup Moonshot launched Kimi K3, the largest-ever open-parameter model yet released, with 2.8 trillion parameters.

Arguably, in line with the distributed security theory of open source, there is one final advantage to embracing the open approach. As countries fine tune and test a model like Kimi K3, build integration tools around it, not only does this spread the use and reliance on Chinese standards, but it also creates free research and development that benefits the underlying systems. Democratizing access to a new standard pours supercharges the innovation cycle around that standard.

Even if "open" is more of a buzzword than a reality, cheap and effective access to the promising tools of the modern economy is an intoxicating promise for Global South countries. How the U.S. — and American companies — choose to compete on this stage could determine power dynamics for many years to come.

Please send feedback, updates and railroad gauge specs to cobun@iapp.org.