when we talk Export control In the tech industry, it’s common for us to think about the restrictions Washington has imposed on the US and China’s access to the most advanced chips. However, Beijing has also had measures in place for years to control the exit of certain technologies from the region. What is interesting is that China now plays the role of not only trying to overcome the obstacles of its rivals: its artificial intelligence models are gaining an international presence. This new situation raises a fundamental question: how to promote this expansion without losing control over certain strategic technologies.
This tension already has a concrete proposal. On October 16, the Chinese Ministry of Commerce and Science and Technology proposed revising the catalog of banned or restricted exports to include 40 technologies related to artificial intelligence, biotechnology and solar energy. Among them, 23 are in the AI field, while five are related to biomedicine and another 12 to the photovoltaic sector. We are not yet talking about definitively approved restrictions: the proposal remains in public consultation, although since the announcement it is necessary to register some technical transfers before it can be carried out.
China wants to protect its technology without closing the door to its international expansion
To understand the scope of these measures, two concepts must be distinguished: one is to prohibit exports and to require approval for its implementation According to the Ministry of Commerce of China, 14 Technology Subject to the contemplated ban on overseas transfers, another 26 would require licences. Furthermore, exporting technology does not necessarily mean shipping an actual product out of the country. This may involve transferring your ownership or granting a license to use overseas, activities covered by registration requirements.
In artificial intelligence, the focus moves beyond models that answer questions or generate text. The proposal covers technologies associated with agents, systems capable of performing tasks with a certain degree of autonomy, and technologies associated with computing clusters, sets of tools that work together to process large workloads. According to the Associated Press, the technology required from models with at least two trillion parameters could be subject to approval. Bans have also been proposed for certain agent technologies for sensitive environments such as defense or critical infrastructure. The Commerce Department justifies part of these measures because of the risk of systems capable of autonomously creating or modifying malicious code.
The key to understanding these restrictions is to distinguish between the technology that allows a tool to be used and the technology it offers. Chinese authorities have clarified that accessing their models through websites, applications, cloud services or APIs, interfaces that allow programs to connect to others, does not in itself constitute an export of the underlying technology. Let’s think about a European company that incorporates a Chinese model via an API: You can use it without transferring that technology to your server. The proposal also excludes knowledge already disseminated through open source or academic publications, while considering simplified procedures for certain new transfers related to open development.
This distinction becomes particularly important if we look at the prominence China’s artificial intelligence developers have achieved. Dipsic and Alibaba are two examples of an industry that has managed to expand its international reach, particularly through models accessible to developers from other countries. According to a report published by Hugging Face in March 2026, models made in China accounted for 41% of downloads on its platform in last year’s analysis. The expansion fits with Beijing’s official policy, which advocates global AI adoption by 2025 and collaboration through open source initiatives.
Concerns about Beijing retaining control over its most advanced powers are also not new. In July 2026, Reuters noted that Chinese authorities had held discussions with companies such as Alibaba, ByteDance and ZI about the possibility of limiting foreign access to some AI models. including some not yet published. Those meetings also discussed possible sanctions for technology leaks and bans on financing national agencies in the sector, the agency suggested. Although these initiatives are made in the context of technological competition with the United States, the Chinese Ministry of Commerce maintains that the review of the catalog is not specifically directed against any country, region or entity.
It remains to be seen how this initiative will translate into concrete obligations for companies. Public consultation will end on November 15, but that date does not coincide with the implementation of the revised catalogue, for which a specific timetable has not yet been announced. Meanwhile, temporary registration requirements will apply for certain technology transfers.
Picture | Fee minutes
In Xataka | The new frontier of AI is trying to predict humans
