China is considering a dramatic reversal of its open-access AI strategy. According to an exclusive Reuters report, the Ministry of Commerce has held meetings over the past month with Alibaba, ByteDance, and the startup Z.ai to discuss restricting overseas access to the country's most advanced artificial intelligence models.
The proposed measures would go beyond a simple export ban. They would also cover open-weight models — the freely downloadable systems that made Chinese AI wildly popular with developers worldwide. Alibaba's Qwen, ByteDance's Doubao, and Z.ai's GLM-5.2 are among the models that could be affected.
Chinese officials have floated several ideas. One would treat the leak or theft of proprietary AI as a national security crime. Another would limit which foreign investors can fund Chinese AI startups. A panel of legal scholars has proposed a tiered system: basic tools would require only a light filing, stronger models would face security reviews, and the most sensitive systems would be locked to domestic use only.
The sources cautioned that nothing has been decided yet. Any curbs might apply only to future models, and no timeline exists.
A Sharp U-Turn
The move marks a sharp departure from China's strategy of winning global goodwill by giving away its models. European developers in particular leaned on cheap, high-quality open-weight models from Chinese firms as an alternative to expensive American systems. Curbing access would thin that supply and could raise costs for businesses worldwide that have built products on top of Chinese AI.
Industry analysts note the symmetry with US policy. Washington has moved aggressively to prevent China from copying American models, and recently restricted Anthropic's frontier systems on national security grounds — the very thing Beijing now fears in reverse. China has already grounded its AI researchers domestically and restricted foreign investment in its tech sector. Treating frontier AI models as state assets would be the next logical brick in that wall.
What's at Stake
Chinese open-weight models have been a gift to the global developer community. They offered near-frontier performance at a fraction of the cost of proprietary Western systems, democratizing access to advanced AI.
If Beijing follows through, the era of freely downloadable Chinese AI could quietly close — reshaping the global AI landscape and potentially accelerating fragmentation of the AI ecosystem along geopolitical lines.
For now, the talks remain exploratory. But the signal is clear: the world's second-largest AI power is reconsidering the open-door policy that made it a favourite among developers everywhere.
