Draft Export Rules Target Overseas Distribution of Domestic Artificial Intelligence Crown Jewels
China’s proposed artificial intelligence export curbs, released on October 16, distinguish between foreign customers using Chinese AI services and foreign companies obtaining the underlying models, data, or technical expertise needed to operate or reproduce them independently, Reuters reported on October 28. The draft regulations add 23 AI-related technologies to a list banning or restricting their export, setting up a regulatory framework to govern the global footprint of domestic tech giants.
The regulatory shift aims to keep valuable intellectual property and research and development onshore while permitting firms to sell access to their technology abroad. This business model matches the path the industry was already taking, according to Alfredo Montufar-Helu, Beijing-based managing director at consultancy Ankura. Under the proposed framework, overseas users could still access Chinese AI through websites and cloud platforms, while actual deployment overseas or the transfer of core technology would face government scrutiny.
Evaluating Technical Thresholds and Compliance Burdens
The draft rules specify three distinct tests to determine whether a model falls under export restrictions. Regulators will evaluate the computing work required to train a model, using a threshold of 10 to the power of 26 calculations. They will also measure model size by at least 2 trillion parameters, or check if the system places among the world’s top 10 in specific assessments of performance, use, or efficiency.
Liu Jing, a senior partner at Beijing Dacheng Law Offices who participated in consultations with policymakers, noted that companies want more clarity regarding the exact tests used to evaluate models. Law firm Han Kun also raised questions about how regulators intend to count the size of advanced models that activate only a fraction of their components for each specific task.
Furthermore, developers are seeking clarification on whether they can train or run their most advanced models on foreign-owned overseas servers. One person briefed on the discussions, who declined to be identified because the talks were private, indicated that in a worst-case scenario, developers might be forced to maintain separate domestic and overseas versions of the same downloadable model.
Corporate Deals and International Infrastructure Challenges
Beyond model releases, the restrictions reach deep into corporate transactions and foreign investment. Registration requirements that took effect immediately on October 16 cover ownership transfers that put Chinese technology under foreign control, including qualifying transactions already underway, according to an analysis by law firm Fangda. Beijing previously demonstrated these enforcement risks earlier in the year by ordering Meta to unwind its more than $2 billion acquisition of Chinese-founded AI startup Manus.
The regulatory pressures arrive as Chinese AI firms and semiconductor startups navigate a broader downturn in local AI and tech stocks. Companies like DeepSeek, Moonshot, and Z.AI currently offer models to international customers via web interfaces and cloud platforms. At the same time, some developers rely on overseas computing facilities to bypass hardware bottlenecks caused by United States export restrictions that block Nvidia’s most advanced processors from entering China.
While the draft provides simplified approval pathways for Chinese companies building and operating their own overseas computing facilities, it does not expressly extend those simplifications to transferring technology to foreign customers for building equivalent systems, according to Han Kun’s analysis. The controls also capture methods for coordinating more than 10,000 AI chips simultaneously.
A Strategic Counterweight to Global Hardware Controls
The emerging oversight acts as a structural counterpart to Western hardware limitations, according to Yang Jingru, a partner at law firm Jincheng Tongda & Neal. Where the United States restricts the inflow of computing power, hardware, and capital into China, Beijing is systematically managing the outflow of models and technological assets.
The public consultation period for the proposals remains open until November 15. China’s commerce ministry did not respond to requests for comment regarding the timeline for final adoption, though sources familiar with the matter anticipate formal implementation by the end of the year.
