China Proposes Export Controls on AI Technologies: What It Means for Global Operators

This revision is not a blanket tightening—it draws precise red lines. For companies serving global users via APIs, SaaS, or app interfaces, most business models fall outside the scope of technology export controls.

What Happened

On October 16, 2026, China's Ministry of Commerce and Ministry of Science and Technology issued Announcement No. 48 of 2026, soliciting public comments on revising the Catalogue of Prohibited and Restricted Export Technologies and simultaneously implementing pre-export registration for affected technologies. The proposal adds 40 key technologies across AI, biomedicine, and photovoltaics—23 of which are AI-related, covering large models, AI agents, computing clusters, and information processing.

Of these, 14 items are proposed as prohibited exports and 26 as restricted. In the AI domain, the focus is on technologies that could be weaponized—such as AI-driven malicious code generation and vulnerability mining—as well as large model weights, source code, architecture designs, and computing cluster scheduling that meet specific thresholds.

Scope: Precise and Limited, Not "All AI"

The controls are not a blanket restriction on all AI technologies. Officials emphasized that the scope is precise and limited, with most items defined by specific technical indicators, and products manufactured using these technologies are not covered.

The following scenarios are explicitly exempted or eligible for facilitated procedures:

  • Open-source or public-domain technologies: Technologies already released via open-source or academic publications are excluded from controls.
  • API calls, SaaS, app/web interactions, lightweight edge deployment: Using restricted large models to provide these services to overseas users does not fall under technology export management.
  • Overseas deployment on self-owned cloud: Falls within scope but qualifies for automatic licensing in low-risk scenarios, effectively "file and go."
  • Self-built overseas computing centers, chip foundry services, joint R&D, participation in international standards: Eligible for automatic licensing and other facilitation measures.

Practical Implications for Global AI Operators

For companies serving global users via APIs, SaaS, or apps, the core message is clear: service-trade models are largely unaffected; it is technology transfer that requires compliance assessment.

The key test is whether restricted technology is transferred from within China to overseas recipients. If model weights, source code, and architecture remain in China, and overseas users only receive outputs via API calls, this is not considered a technology export. However, transferring technology ownership, licensing overseas use, or changing ownership/control of domestic technology to foreign entities requires pre-export registration through the Ministry of Commerce's unified platform.

Computing Clusters in Focus

The inclusion of computing cluster technologies warrants attention from infrastructure operators. Controlled items cover high-speed interconnects, distributed training fault tolerance, efficient communication scheduling, and heterogeneous cluster unified scheduling for large-scale intelligent computing clusters (defined as those with over 10,000 AI accelerators).

This means that cross-border transfer of computing cluster technology itself requires compliance review, but providing cloud-based computing services or model API access is not equivalent to a technology export.

Transitional Arrangements

From the announcement date until the revised Catalogue formally takes effect, exporters must register covered technology exports on the Ministry of Commerce's unified platform. Ongoing but incomplete exports are also subject to registration.

Public comments can be submitted via email, fax, or mail between October 16 and November 15, 2026.

Conclusion

This revision is less about "closing doors" and more about "drawing lines." For API- and SaaS-first operators, the compliance boundary is clearer than ever: as long as core technology stays within China, the path for service-based expansion remains open. The real challenge is building internal compliance awareness—distinguishing which business models involve technology transfer and which are purely service delivery.

In an era of tightening global AI governance, compliance capability itself is becoming essential infrastructure for going global.

Last updated October 17, 2026

Where to go next