DeepSeek and MiniMax Win Approval Under China AI Export Controls for Overseas Deployment

DeepSeek and MiniMax Secure State Approval for Overseas AI Deployment

DeepSeek and MiniMax have successfully completed registration procedures and obtained approval for AI global use under China's newly proposed regulatory framework1

1

. According to Yuyuan Tantian, a WeChat account affiliated with state broadcaster CCTV, the two leading Chinese AI companies secured open-source and overseas licensing rights following the announcement of proposed export controls on AI technologies3

3

. The timing proves significant as DeepSeek reportedly raised $15bn from Tencent and CATL this week1

1

. This approval does not mark the first time these models became available internationally—users worldwide were already accessing DeepSeek's services2

2

. The distinction lies in the formal authorization to transfer, license, and deploy underlying AI technology overseas, separate from merely providing access through online services.

Understanding the Proposed Export Controls on AI Technologies

On October 16, China's commerce and science ministries proposed adding 23 AI-related technologies to the country's export restriction list1

1

. The proposed export controls on AI technologies cover large language models, AI agents, computing clusters, AI chip designs, and high-speed links for large computing infrastructures4

4

. Companies must register before exporting any listed technology, with public comments accepted until November 151

1

. The regulatory framework targets three specific areas to regulate AI technology transfers. First, data security: the rules would ban exports of large models trained on China's "core data" and restrict models trained on "important data" regardless of size1

1

. Second, sensitive applications: AI agents above certain performance levels face restrictions, with outright bans on agents designed for critical infrastructure, defense, or core government operations1

1

. Third, computing power: the catalogue sets specific thresholds for large models, including training compute above 10^26 floating-point operations, more than 2 trillion parameters, or capability ranking in the global top ten1

1

.

What Remains Outside the China AI Export Controls

Despite national security concerns driving the policy, most AI technologies will remain freely exportable1

1

. Critically, offering restricted AI models for overseas use through APIs, apps, web pages, SaaS, or lightweight local deployment does not constitute a technology export under the proposed rules1

1

. "Simply put, AI companies can continue with their commercialization efforts as usual," the CCTV-affiliated post stated1

1

. Open-source technology receives automatic licensing through an "immediate reporting and export" process1

1

. Chinese companies operating their own computing centers abroad, chip tape-out overseas, joint research projects, and work on international standards also qualify for simplified procedures1

1

. Technology already in the public domain through open source releases or academic papers falls entirely outside the restrictions1

1

. This distinction matters because Chinese open-weight models have become vital resources for developers and research institutions seeking alternatives to U.S. AI systems2

2

.

The Geopolitical Context Behind Beijing's AI Strategy

Beijing's approach contrasts sharply with Washington's strategy. The CCTV-affiliated account cited the U.S. Commerce Department's AI diffusion framework from January 2025 and an "alliance strategy" attributed to Anthropic's CEO Dario Amodei1

1

. "But China did not take this path," the post emphasized1

1

. The proposed controls represent Beijing's latest move to limit transfers of tools it views as vital to national security amid intensifying geopolitical competition with the United States3

3

. Chinese AI companies have gained global traction by offering relatively affordable models that developers can download, customize, and integrate into their applications, positioning Chinese models as alternatives to proprietary systems from OpenAI and Anthropic2

2

. As AI becomes strategically important, Beijing seeks greater oversight over how underlying technology moves across borders while maintaining commercial momentum2

2

.

Implications for Global Developers and AI Deployment

For ordinary users, there is little indication of losing access to these models2

2

. The clearance suggests Beijing wants leading Chinese AI developers to continue expanding abroad despite new safeguards4

4

. However, the rules create a two-tier system: accessing models through online services remains straightforward, while exporting, licensing, or deploying underlying technology overseas may involve additional regulatory requirements2

2

. Liu Jing and Liu Hongmei of Beijing Dacheng Law Offices participated in consultations, noting that control points for some AI items changed after the first draft and many company concerns appear in the published version1

1

. Liu Hongmei confirmed that restricted technologies can still go abroad with proper licensing, which should not disrupt companies' overseas plans1

1

. The move may ease concerns that proposed controls would cut off overseas access to Chinese open-weight models, signaling Beijing's intent to separate routine commercial deployment from exports of sensitive AI capabilities4

4

. With draft rules open for public comment until November 15, final provisions could still alter compliance burdens for Chinese AI developers operating globally4

4

.