Beijing weighs limiting foreign access to AI models from Alibaba and ByteDance

China is considering restrictions on foreign access to the country's leading AI models. In June–July 2026, the Ministry of Commerce held a series of closed-door meetings with the largest tech companies, Reuters reports.

Who took part in the meetings

Representatives of Alibaba, ByteDance and the startup Z.ai took part in the talks. The meetings ran for about a month and focused on a single question: whether, and how exactly, to restrict foreign users' access to the most powerful Chinese-made AI systems.

Key facts from the Reuters report:

• The talks were initiated by China's Ministry of Commerce
• Participants: Alibaba, ByteDance and the startup Z.ai
• The discussions took place approximately in June–July 2026
• The restrictions under consideration apply to foreign users; the domestic market is not affected

Why open Chinese models matter to developers worldwide

Over the past two years, China has steadily released powerful open-source AI models that have become a significant resource for teams around the world. Alibaba develops the Qwen family, models that regularly rank high on international benchmarks and are available for free. Offerings from ByteDance and the startup Z.ai are also widely used outside China.

For small companies and independent developers, especially in regions with limited budgets such as Southeast Asia, Africa and Latin America, these tools have become a real and affordable alternative to paid APIs from OpenAI, Google and Anthropic. Thousands of products have already been built on open Chinese models. If Beijing introduces restrictions, teams will have to find replacements or operate under new legal risks.

A mirror response to US export controls

The discussion of restrictions comes against the backdrop of years of technological rivalry between the US and China. Since 2022, Washington has steadily tightened export controls: it has banned shipments of advanced chips to China, restricted access to semiconductor manufacturing equipment and blocked American investors from taking part in Chinese AI projects.

Mirror measures on AI models fit the same logic: creating a symmetrical lever of pressure. They would protect the commercial interests of Chinese developers, reduce the risk of leaks of strategic AI technology and give Beijing a new tool in geopolitical bargaining with the West.

Notably, the talks involved not only corporate giants, Alibaba and ByteDance, but also the startup Z.ai. This suggests an intent to cover the full range of companies with models popular abroad, not just the systemically important players.

What this means

The mere fact that the Ministry of Commerce is discussing such measures is a signal to the market: guaranteed, free access to open Chinese AI models can no longer be taken for granted. Developers who have built products on these systems should already be thinking about diversifying their providers. The global market for AI tools risks becoming even more fragmented, with some models available only in the West and others only in China.

Frequently asked questions

Which models could be affected by the restrictions?

Reuters does not name specific systems; the discussion was general. The participants include Alibaba with its Qwen family, ByteDance and the startup Z.ai, whose models are widely used on open platforms abroad.

When might concrete decisions emerge?

No firm timeline has been announced. According to Reuters, officials are still discussing positions rather than having made decisions. The Ministry of Commerce is gathering input from market participants before drafting specific proposals.

Why was the startup Z.ai involved in the talks, and not other companies?

Reuters does not explain the criteria for choosing participants. However, including a startup alongside large corporations indicates that the authorities are looking at the market as a whole: not only the giants but also promising smaller developers with an international presence.