First Movers Clear Beijing’s New AI Export Hurdle
BEIJING, October 23, 2026 — DeepSeek and MiniMax have become the first artificial intelligence firms to complete registration under China’s newly proposed export control framework. The two companies secured approval for the global deployment and open-sourcing of their models, a move that signals compliance with Beijing’s latest regulatory measures.
The clearances were confirmed by Yuyuan Tantian, a self-media outlet owned by state broadcaster CCTV, on Thursday. In a separate post on Friday, the account noted that overseas internet users had urged other Chinese open-source large language models to follow suit. DeepSeek and MiniMax did not immediately reply to requests for comment regarding the specific details of their registration.
This development occurs against the backdrop of Beijing’s introduction of proposed export controls on October 16. The framework covers 23 distinct AI technologies, including models, agents, and computing clusters. It represents the Chinese government’s latest effort to limit the transfer of tools it views as vital to national security amid rising geopolitical competition with Washington.
Defining the Red Lines for AI Technology
Understanding the new rules requires looking past the headline of “export controls.” The Chinese Ministry of Commerce and Ministry of Science and Technology opened a public comment period on October 16 for a revised “Catalog of Technologies Prohibited or Restricted from Export.” The revision is described as routine work conducted in accordance with laws and regulations, aiming to regulate companies’ technology export activities rather than the technology itself.
In plain terms, the catalog draws a sharp line between what is allowed and what is not. Large models trained on “core data” and offered as services abroad would be prohibited from export. Models trained on “important data” offered as services would be restricted. AI agent technologies meeting certain performance levels would be restricted, escalating to a prohibition if used in highly sensitive scenarios such as critical information infrastructure, national defense, or core government affairs.
However, the practical upshot for most developers is that commercialization activities remain largely untouched. Providing API access, app or web interaction, SaaS applications, and lightweight on-device deployment fall outside the scope of large-model technology export management. The catalog also sets quantitative thresholds to reduce the burden of judgment on companies, including computing volume exceeding 10² floating-point operations and parameter counts above 2 trillion.
How Beijing’s Approach Differs from Washington’s
The regulatory strategy in China contrasts with the approaches seen in the United States. The U.S. Commerce Department published the “AI Diffusion Framework” in January 2025, which was designed to manage advanced computing chips and some model weights based on the sensitivity of the destination. Another U.S. approach involves bloc-based circles, with Anthropic’s chief executive having proposed a “coalition strategy” in which a small group of countries would jointly control the AI technology supply chain.
By contrast, China’s revision emphasizes routine, rules-based management. The Ministry of Commerce’s accompanying Q&A devoted noticeably more space to explaining what is not covered than what is covered. This detail is widely seen as signaling that the policy’s core is drawing boundaries rather than simply restricting outbound AI technology. The goal is to ensure that the vast majority of AI technologies remain exportable under simplified licensing frameworks, particularly for open-source developments and firms operating overseas computing centers.
Lawyers at Beijing Dacheng Law Offices, including Liu Jing and Liu Hongmei, noted that existing overseas footprints and business plans will not be disrupted. They explained that while prohibited technologies cannot be exported, restricted technologies can still be licensed. This distinction is crucial for companies that have already established a presence abroad.
Implications for Markets and Industry Compliance
For the industry, the registration completed by DeepSeek and MiniMax provides a replicable template for peers. It demonstrates that leading players can clear compliance procedures on both the open-source and overseas licensing paths. This is particularly significant for MiniMax, a Hong Kong-listed entity (00100.HK), which gains greater compliance certainty for its overseas business. This clarity helps ease investor concerns about policy uncertainty, although the catalog is still at the public-comment stage.
The regulatory focus on export activities rather than the technology itself means that normal commercialization pace will not be disrupted for most firms. The real constraints concern technology flows involving core data, critical infrastructure, and highly sensitive scenarios. These factors will shape how leading companies choose their deployment strategies in the coming months.
China has been developing its national AI agenda since 2006, aiming to become a global leader by 2030 as outlined in its 13th Five-Year Plan. With the addition of AI-related entries to the export catalog, the government is filling a gap in its technology import and export management system. The catalog has been adjusted and revised multiple times since 2001, and this latest update reflects the rapid evolution of the AI sector. As the public comment period concludes, the focus will shift to implementation details, approval timelines, and licensing standards, which remain key variables for AI-sector valuations.
Sources
- China clears DeepSeek and MiniMax for global AI deployment
- China's DeepSeek, MiniMax win approval for AI global use, state media-affiliated entity says
- DeepSeek, MiniMax First to Complete AI Technology Export Registration as China's 23-Item Tech Catalog Moves Toward Implementation — BigGo Finance
Photo: Solen Feyissa / Unsplash (Unsplash License)
