China plans to tighten export controls on open-weight models and 23 key AI technologies
The move targets frontier AI models and 23 key technologies, with restrictions based on training data and system capabilities. The draft rule is part of broader efforts to regulate AI exports.
China's commerce and science ministries are advancing a draft rule to tighten export controls on open-weight models and 23 key AI technologies. The regulation aims to restrict the export of AI systems trained on sensitive Chinese data or those exceeding specific thresholds, such as 2 trillion parameters or 10^26 FLOPs of training compute. The initiative reflects growing concerns over the potential misuse of advanced AI capabilities and the need to safeguard national security interests.
The proposed export controls focus on 23 key areas, including frontier models, AI agents, and autonomous malware tools. These restrictions are designed to prevent the uncontrolled spread of AI technologies that could be used for malicious purposes. The draft rule is part of a broader regulatory framework aimed at ensuring that AI development aligns with China's strategic and security objectives.
The regulation introduces specific thresholds for export restrictions. Systems trained on sensitive Chinese data or those with more than 2 trillion parameters, 10^26 FLOPs of training compute, or a global top-10 ranking will face export limitations. These criteria are intended to identify and control the export of highly capable AI models that could pose risks if deployed abroad.
The tightening of export controls is likely to increase compliance costs for companies involved in AI development and export. It may also lead to increased vendor lock-in as firms seek to navigate the new regulatory landscape. Additionally, the move could influence global governance of AI, prompting other nations to reconsider their own export policies and regulatory approaches.
The regulation is still under development, with the draft rule signaling a significant shift in China's approach to AI governance. As the rule evolves, it will be crucial to monitor how it interacts with existing international frameworks and how it affects the global AI ecosystem. The long-term implications of these export controls remain to be seen, but they are expected to shape the trajectory of AI development and deployment worldwide.