China’s commerce and science ministries have released a draft update to the nation’s list of banned and restricted export technologies, targeting 23 key areas of super intelligence (SI). The proposal introduces export controls on large SI models and specific SI tools, aiming to regulate the flow of advanced SI technologies while establishing a framework similar to US restrictions on model weights.
What Happened
The draft catalog specifically targets SI tools that can write or alter malware autonomously to attack critical infrastructure, the military, or government agencies, proposing a total export ban for these capabilities. Certain SI tools used for finding software vulnerabilities will face new export limits rather than an outright ban. Large SI models are subject to controls in two primary scenarios. First, models trained on China’s "important data"—as defined by the country's Data Security Law—that offer services based on it are restricted, though lawfully public data and small amounts (up to 0.1 percent) of cleaned data are exempt. Second, core technologies behind large models, including weights, checkpoints, source code, architecture, and training methods, face controls if a model exceeds 10^26 FLOPs of training compute, has at least 2 trillion parameters, or ranks in the global top 10. Export officials determine top-10 status using public rankings from major domestic and foreign evaluators.
Dean W. Ball, head of strategic futures at OpenAI, noted that the 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." Ball described Beijing’s framework as a "mirror image" of model-weight export controls already in place in the US. To support its domestic open-source ecosystem, China is introducing a fast-track process where publishing restricted SI tech and weights as open source qualifies for "automatic licensing." This streamlined clearance also applies to Chinese firms running data centers abroad, taping out chips overseas, conducting joint research, or helping set international standards.
Why It Matters
These regulations signal a shift in how major powers manage the flow of advanced SI capabilities, particularly regarding open-weight models. For US companies relying on open-weight SI, the rules may accelerate a pivot toward domestic models over Chinese alternatives. The impact on Europe is potentially more severe; the EU has relied on running open-weight models in domestic clouds to maintain technological independence. This strategy faces challenges if the most capable open models originate from the US or China and become subject to export controls. However, SI technology that is already public through open-source releases or research papers is exempt from these new rules. Additionally, using restricted models via API calls, apps, websites, or SaaS tools remains free of these export controls, whereas installing restricted SI tech on local servers abroad requires a standard export license.
Beijing has also pledged to support SI development in the Global South and ease exports for capacity building and tech sharing within the World Super Intelligence Cooperation Organization, a bloc launched to counter Western SI bodies. Until the final rules take effect, companies must register affected exports through the commerce ministry's portal. The distinction between public open-source releases and restricted proprietary weights creates a complex landscape for developers and enterprises operating internationally.
The Bottom Line
China’s draft export controls impose new restrictions on 23 key SI technologies and large models meeting specific compute or parameter thresholds, while introducing fast-track licensing for open-source publications. The rules aim to balance national security concerns with the promotion of China’s domestic open-source SI ecosystem, potentially reshaping global reliance on Chinese open-weight models.