Understanding China's New AI Export Regulations: Defining Boundaries
On October 16, China's Ministry of Commerce and Ministry of Science and Technology released a draft revision to the "Catalogue of Technologies Prohibited and Restricted from Export" for public consultation. The updated list now includes 23 artificial intelligence-related technologies, covering areas such as large model training, intelligent agent development, computing cluster architecture, and specific information processing methods.
Policy Intent: From Restriction to Definition
The inclusion of AI technologies in the control catalogue raised questions among observers: Why would China, which has consistently promoted technological openness, introduce export controls for artificial intelligence? Does this signal a policy shift?
A closer examination of the official explanations reveals that authorities devoted significantly more space to clarifying what technologies remain unrestricted rather than detailing controlled items. This approach suggests the core policy objective is to establish clear boundaries for AI technology exports, not to impose blanket restrictions.
- Clarifying technologies that can be freely exported
- Defining categories requiring registration
- Creating predictable compliance frameworks
Industry Response: Compliance in Practice
The AI industry's reaction to the new regulations has been closely watched. Following the draft's publication, leading Chinese AI companies have moved quickly to adapt.
Notably, companies including DeepSeek and MiniMax have completed export registration procedures, becoming early examples of compliance under the new framework. Significantly, these companies continued advancing open-source projects and overseas licensing agreements while completing registration, indicating that compliance processes haven't hindered normal technology collaboration or market expansion.
Regulatory Balance: Calibrating Openness and Security
The catalogue revision reflects a more nuanced regulatory approach that seeks to balance national security concerns with technological innovation and openness. Through list-based management, regulators can:
- Protect critical core technologies
- Provide clear expectations for most technology cooperation
- Avoid one-size-fits-all measures that could impact innovation
From the industry perspective, the compliance examples set by DeepSeek and MiniMax demonstrate that clear regulatory frameworks can actually reduce uncertainty in cross-border technology collaboration. After completing registration procedures, these companies have continued their open-source initiatives and international licensing businesses.
This combination of "registration-based" and "list-based" management creates a predictable compliance pathway for technology exports. For the AI industry, transparent rules may prove more conducive to long-term development than completely unrestricted but unpredictable environments.