China has proposed expanding its technology export controls to cover 23 artificial intelligence technologies, potentially putting large models, AI agents and computing clusters under tighter government oversight while leaving several common forms of online AI access outside the regime.
China’s Ministry of Commerce and Ministry of Science and Technology announced on October 16 that they want to add 40 technologies across artificial intelligence, biomedicine and photovoltaics to the country’s Catalogue of Technologies Prohibited or Restricted from Export. Twenty-three are AI technologies.
Reuters described the proposal as Beijing’s latest move to tighten oversight of technologies it considers important to national security, amid intensifying technological competition with the United States.
Models, agents and computing clusters are included
The Chinese government says the AI portion covers technologies related to large models, intelligent agents, computing clusters and information processing.
Across all three sectors, the government proposes classifying 14 technologies as prohibited from export and 26 as restricted. Examples include AI technology capable of autonomously generating or modifying malicious code and AI-assisted vulnerability discovery.
The public consultation remains open until November 15, 2026.
However, one part of the system is already active.
Until the catalogue is formally revised, exporters transferring technologies covered by the announcement must complete pre-export registration with the Ministry of Commerce. The requirement also applies to relevant exports that have started but have not yet been completed.
APIs are treated differently from overseas model deployment
The most important distinction for AI companies concerns how technology crosses borders.
China’s commerce ministry said that overseas local deployment of a model containing restricted technology can fall under technology-export rules.
But offering access through APIs, apps, websites or software-as-a-service platforms does not count as exporting the underlying large-model technology.
Reuters noted that this distinction could allow Chinese developers such as DeepSeek and Alibaba to continue serving international users through online services while Beijing retains greater control over transfers involving the technology itself.
That distinction could become commercially significant.
A customer accessing a Chinese model through an API may face a very different regulatory situation from an enterprise seeking to deploy the same underlying technology on infrastructure outside China.
Open-source technology receives another carve-out
The proposal also excludes technologies that have already entered the public domain through open-source release or publication in academic papers.
That creates an unusual strategic boundary.
For AI labs, decisions about whether to release model technology openly, provide it only through an API or support private overseas deployments may no longer be purely technical or commercial decisions. They can affect how export rules apply.
China says the revision is part of normal technology-trade management and does not target any specific country, region or entity.
For founders and infrastructure teams, the larger message is clear.
AI deployment architecture is increasingly becoming geopolitical architecture.
Where a model runs, how customers access it and whether its underlying technology is openly published can now determine not only performance and cost—but whether the technology can cross a national border at all.
