China's commerce and science ministries announced plans to tighten export controls on 23 key AI technologies, including open-weight models with over 2 trillion parameters. The draft rule targets areas such as frontier models, AI agents, and autonomous malware tools, which could face export bans if they can write or alter malware to attack critical infrastructure, the military, or government agencies.
AI tools for finding software vulnerabilities will face new export limits, though they will not be banned outright.
Large AI models also face export controls in two cases: if trained on China's 'important data' or if they hit 10^26 FLOPs of training compute, have at least 2 trillion parameters, or rank in the global top-10.
Export officials will use public rankings from major domestic and foreign evaluators to determine top-10 status.
"This move does not mean 'China is shutting down open-weight models,' but it does indicate the Chinese government establishing the optionality to shut down future open-weight models at will," said Dean W. Ball, head of strategic futures at OpenAI. Ball added that Beijing's framework is a 'mirror image' of the model-weight export controls the US government has already put in place.
The announcement follows a broader strategy to protect China's domestic AI development while managing export risks. To avoid choking off its open-source boom, China is introducing a fast-track process for publishing restricted AI tech as open source, which qualifies for automatic licensing.
This allows companies to use restricted models through API calls, apps, websites, SaaS tools, or lightweight versions on local devices without export controls.
China also pledged to support AI development in the Global South and ease exports for capacity building, tech sharing, and joint rulemaking within the World Artificial Intelligence Cooperation Organization. Until the final rules take effect, companies must register affected exports through the commerce ministry's portal.
Source: thedecoder
