DeepSeek and MiniMax have signed up under Beijing's draft artificial intelligence export rules, according to Yuyuan Tantian, a CCTV-affiliated account that reported the registrations on Thursday. The two Chinese developers have also completed open-sourcing their models and licensing their use overseas, the account said. Under the framework the account describes, both steps fall within automatic licensing, placing the two developers on the lightest compliance track the draft offers rather than in case-by-case reviews.

In a separate post on Friday, Yuyuan Tantian said overseas internet users have urged other Chinese open-source model developers to follow suit.

On October 16, China's Ministry of Commerce and Ministry of Science and Technology announced the opening of public comment on a revised Catalogue of Technologies Prohibited and Restricted from Export. The draft lists 23 AI technologies, including large models, agents, compute clusters and information processing. The move raised questions in markets and foreign policy circles about whether China was closing the door on AI exports. According to the account, the Commerce Ministry's Q&A devoted more words to what the rules do not cover than to what they do. Because the catalogue remains in its comment stage, the measures have not yet taken effect; until then, the existing catalogue continues to govern technology exports.

Restricted items can still win licenses, lawyer says

Lawyers Liu Jing and Liu Hongmei of Beijing Dacheng Law Firm took part in consultations on the draft, which began in the first half of the year. They told Yuyuan Tantian that control points for some AI entries kept shifting with the industry as successive drafts moved forward. Many of the concerns companies raised at those meetings were reflected in the draft, and their bigger question now is whether clearer operational guidance will follow, the account said. For companies weighing overseas deployments, the practical watch items are the final catalogue text and any follow-up guidance spelling out how thresholds and licensing reviews will operate in practice.

Under the framework described by Yuyuan Tantian, the following do not constitute an export of a model: providing API access, interaction through an app or website, SaaS services, or lightweight on-device deployment with a restricted model. For overseas users, that reading means running Chinese open-source models through an API, app, website or SaaS subscription would not itself count as an export under the draft.

The account said most artificial intelligence (AI) technology will continue to be exported freely once the measures take effect. Open-sourcing AI, running self-built compute centers abroad, taping out chips overseas, joint research and development with foreign resources, and work on international standards all qualify for automatic licensing — close to file-and-go. Members of the World Artificial Intelligence Cooperation Organization will receive stronger support for building capabilities and shaping rules together.

The draft delineates three areas. A model trained on core national data and provided as a service would be barred from export, while a model trained on important data would be restricted, irrespective of size or number of parameters. Agent technology above a certain level of performance would be restricted; if it is mainly intended for critical information infrastructure, defense or core government systems, it moves up to prohibited. The third category is compute, covering the design of AI chips and the high-speed interconnects that link nodes inside large AI compute clusters. Illustrative numeric thresholds include training compute above 10^26 floating-point operations, more than 2 trillion parameters, or capability ranked in the global top ten.

As Cryptopolitan reported in May, DeepSeek's flagship V4-Pro has 1.6 trillion parameters. That places the flagship below the draft's illustrative 2-trillion-parameter threshold, although the catalogue's data-source and end-use tests can restrict or bar a model irrespective of its size.

Liu Hongmei said that once the rules come into effect, technology in the banned category cannot be exported. She added that companies could seek licenses for restricted items without having to scrap existing overseas plans.

Yuyuan Tantian calls the revision routine, citing changes since 2001

The account's reporting also addressed the international backdrop. The U.S. Department of Homeland Security has instructed federal civilian agencies to discontinue use of DeepSeek products and any third-party tools developed on its models, Cryptopolitan reported; that order was signed by Secretary Markwayne Mullin on July 24, 2026. Yuyuan Tantian said Washington categorizes AI export restrictions by country and alliance pointing to the January 2025 AI Diffusion Framework.

By contrast, the account said, China has revised the catalogue several times since 2001, and the current revision is routine work under existing law. Oversight, it said, falls on the act of exporting technology rather than on the technology itself.