The U-Turn: China May Stop Giving Its Best AI Away — And That Changes Everything

Beijing is quietly weighing whether to restrict the open models that became its greatest strategic weapon. If it does, the entire logic of the AI race inverts overnight

The Reversal Nobody Expected

For weeks, watchchina.ai has documented a single, consistent thesis: China's open-weight strategy — giving its best AI away for free, to the entire world — is its most powerful weapon. It drove the 140-trillion-token explosion, the 45% share of global OpenRouter traffic, the diffusion across the Global South that no export control could touch. Today, that entire strategy was thrown into question by Beijing itself.

Chinese authorities have held talks with Alibaba, ByteDance, and Z.ai about whether to restrict foreign access to their most advanced models, including ones not yet released, Reuters reported today, citing three people familiar with the discussions. Nothing has been decided, and the ministries involved have made no official comment — but officials have gone as far as sketching options, including a bar on public release or a limit to domestic use only. As TIME framed it with precision: Chinese AI companies have made inroads globally by giving their models away for free — and now Beijing is weighing whether to stop them. That would be a dramatic U-turn from the open-weight approach that made Chinese AI a global force in the first place.

Why Would China Abandon What's Working?

The apparent paradox dissolves once the strategic logic is examined — and it reveals how completely the calculus has shifted. There are three plausible drivers, each pointing to a different conclusion about what Beijing now believes.

The first is that openness has served its purpose. If the goal was to establish Chinese models as global standards and build ecosystem lock-in — 100,000 Qwen derivatives, adoption across Southeast Asia and the Gulf — then that beachhead may now be secure enough that Beijing can begin converting influence into control and revenue.

The second is security: as watchchina.ai reported, Anthropic has accused Alibaba of distillation, and Washington increasingly frames open Chinese models as national security threats. Beijing may calculate that its most advanced models are now strategic assets too valuable to hand to rivals who could distill, study, or weaponise them.

The third, and most telling, is confidence. Restricting access to your best models only makes sense if you believe those models are good enough that the world will want them badly enough to accept restrictions — the exact posture the United States adopted, and that China spent months mocking.

The Great Convergence — And What It Means

The deepest significance of this potential U-turn is what it reveals about the trajectory of the entire AI race: the two superpowers may be converging on the same playbook.

For a year, the defining contrast was openness versus restriction — America walling off its best models, China giving its away. watchchina.ai built much of its analysis on that very distinction. If Beijing now moves to restrict its frontier models, that contrast collapses, and both powers end up in the same place: treating top-tier AI as a controlled strategic asset, released freely only at the "good enough" tier while the true frontier stays locked down.

This would be a vindication of a subtler point watchchina.ai has raised repeatedly — that the "China is open, America is closed" narrative was always too simple, and that as Chinese labs approached the frontier, their commercial and strategic incentives to close up would grow exactly as they did for OpenAI and Anthropic. The timing sharpens everything: this debate surfaces just as China prepares to host the World AI Conference on Global AI Governance in Shanghai, where it has positioned itself as the champion of open, inclusive, accessible AI.

A move to restrict its best models in the same month would expose the gap between China's open-AI rhetoric and its strategic self-interest — the same gap the West has been accused of for years. Nothing is decided yet. But the mere fact that Beijing is considering it tells watchchina.ai readers something important: the age of freely-given Chinese frontier AI may have a shelf life, and the window to understand China's open era may be closing even as it peaks.

Sources: Reuters, TIME, CNBC, Xinhua, South China Morning Post