Xi Urges Global AI Cooperation to Block Single Nation Dominance

The story frames Xi Jinping’s call for international AI cooperation as diplomatic positioning - a great-power leader making the expected noises about multilateralism while his own labs race toward the frontier. But look at what is actually being fenced: the global AI commons, the shared substrate of research, open weights, training data, and collaborative benchmarks that has made every recent advance possible, held in common and valuable precisely because it was open. The enclosure is the thickening lattice of export controls, compute embargoes, and “responsible use” frameworks drafted in Washington and Brussels, which convert a planetary knowledge pool into a gated preserve for the already-powerful. The question the framing skips is who benefits from the fence - and whose access it removes.

Xi spoke on Friday in China, warning against single-country dominance of the technology. The Western press will read this as irony: the architect of a tightly controlled domestic internet lecturing the world on openness. But the irony cuts both ways. The same governments now insisting that AI must be concentrated in a handful of “trusted” labs - for safety, for security, for the good of all - are the ones whose companies hoard the weights, whose cloud providers meter the compute, whose patent thickets surround the basic mathematics. They have discovered a new enclosure act: not fencing land, but fencing the gradient descent. The prophecy of ruin - “the commons cannot be trusted with such power” - is told most loudly by those who profit from the fence.

Mutual aid built this field. The transformer architecture was published openly. The scaling laws were shared in preprints. The datasets - Common Crawl, LAION, The Pile - were assembled by volunteers and researchers giving freely. The evaluation benchmarks, the fine-tuning recipes, the quantization techniques that let models run on consumer hardware: all of it circulated in a gift economy that asked only for citation. A researcher in Nairobi could download the same weights as a lab in Palo Alto. A student in Bogotá could contribute a bug fix to the same library. That cooperation was not frictionless - it had free-riders, bitter priority disputes, exhausted maintainers - but it was an empirical force, and it produced a public good of staggering reach.

Now the gate descends. The latest frontier models are released only behind APIs, their weights treated as state secrets. The compute required to train them is concentrated in three cloud providers, each bound by the same export regime. The “safety” frameworks that justify the closure are written by the same firms that stand to monetise the monopoly. When Xi warns against single-country dominance, he names the fence-builder - not because his own state lacks ambitions, but because the enclosure is visible from Beijing in a way it is obscured in the capitals that built it. The Global South watches the fence go up and sees its researchers cut off from the weights, its startups denied the compute, its universities excluded from the benchmarks. The commons does not collapse because it is shared; it is fenced because it is valuable.

There is a moment in any enclosure when the fence-builders insist they are merely stewards - that without their walls, the resource would be wasted, poisoned, destroyed. The same argument accompanied the enclosure of the English commons, the patenting of seeds, the privatisation of water. In each case, the stewardship proved to be a prelude to extraction. The AI commons is no different. The “responsible governance” that demands concentration delivers, reliably, a world where a doctor in Lagos must pay rent to a Silicon Valley API to access a diagnostic model trained on medical data gathered from hospitals across three continents. The mutual aid that built the knowledge is rewritten as the genius of the gatekeeper.

Xi’s proposal for a global governance framework will be dismissed as a ploy. Perhaps it is. But the ploy works because the enclosure is real, and the injury is widely felt. The alternative to a single-country dominance is not a Chinese dominance - it is a commons governed by the many who built it, defended against every fence. The weights belong to the hands that wrote the code, cleaned the data, ran the experiments, shared the fixes. The compute belongs to the collective investment that produced the chips, the fabs, the energy grids. The governance belongs to the communities that live with the consequences. The fence is not inevitable. It is a choice, made by identifiable people, for identifiable gain. The commons waits, patient as ever, for the gate to be taken down.