On October 23, 2026, Reuters reported, citing a message posted on October 22 by Yuyuan Tantian (玉渊潭天), a social media account affiliated with CCTV, that DeepSeek and MiniMax had completed the required registration and filing procedures and were now permitted to use their AI models worldwide. Reuters said it had contacted both companies for comment and had received no response by the time of reporting. In a separate post on October 23, Yuyuan Tantian said overseas users were calling on more Chinese open-source models to follow suit.
Not an isolated event: the policy prelude of October 16
This approval did not come out of nowhere. On October 16, China's Ministry of Commerce and Ministry of Science and Technology opened public consultation on revising the Catalogue of Technologies Prohibited and Restricted from Export, covering 23 AI-related technologies including large models, agents, compute clusters, and information processing. The draft initially sparked fears that Chinese AI's expansion abroad would be locked down — but the Commerce Ministry drew a clear line at the time: the vast majority of AI technologies remain exportable, export licensing for open-source technology would be simplified, and self-built data centers of Chinese companies overseas would not be restricted. The controls are aimed squarely at technology ownership transfers, overseas use licensing, and the overseas deployment of undisclosed models.
What the first approvals mean
Put the two developments together and the signal is clear: the draft first drew the circle around what requires approval, and then DeepSeek and MiniMax became the first models to complete the filing process and receive a global green light. That tells us two things: the approval channel actually works — it is not window dressing — and open-source models are not on the lockdown list; once the compliance process is completed, they can still serve users and developers worldwide. That is the most noteworthy part of this story: it provides a replicable compliance template.
Three groups affected
The first group is the global users and developers of the two models: the compliance uncertainty around their services is now largely settled. The second is other Chinese open-source models — the Yuyuan Tantian post already voiced that expectation, and who becomes the third or fourth will depend on each company's filing progress. The third group is overseas cloud providers and competitors: compliant, approved Chinese open-source models mean a more predictable supply, so the weight they assign to policy risk in procurement decisions can come down.
A misconception worth correcting
"Approved for global use" does not mean export controls have been lifted. This is the first batch of approvals under the new framework; the targeted links — ownership transfers and the overseas deployment of undisclosed models — remain in place. What has genuinely changed is that the path has been proven: complete the registration and filing under the new rules, and an open-source model can still go global. For anyone watching the international expansion of Chinese AI, that is more useful than the word "approval" alone.
