China is preparing new export restrictions on artificial intelligence technology that would allow foreign customers to use Chinese AI services while giving Beijing greater control over the underlying models, technical expertise and infrastructure. The proposed rules reflect an increasingly important strategy in the US-China technology rivalry: expanding international access to Chinese AI without surrendering ownership of its most valuable technological assets.
Released on October 16, the draft regulations distinguish between overseas customers accessing AI through websites and cloud platforms and foreign companies acquiring the technology needed to operate, reproduce or develop equivalent systems independently.
The proposals would add 23 AI-related technologies to China’s list of technologies banned or restricted from export. These include advanced AI models, autonomous agents capable of performing tasks with limited human supervision, and technologies used to design semiconductors and operate large computing facilities.
China is expected to adopt the proposals by the end of 2026, with implementation guidelines to follow, according to a source familiar with the discussions. The draft remains open for public comment until November 15.
Stay ahead of the geopolitical week.
MD Briefing delivers expert analysis across five global fronts — the Indo-Pacific, energy, geoeconomics, European security, and the Middle East — every Monday morning. Free.
The distinction between access and ownership could reshape how Chinese AI companies expand internationally, particularly as restrictions imposed by Washington limit China’s access to advanced computing hardware and semiconductor technology.
Why Beijing Wants to Keep Its AI Technology at Home
China’s proposed framework favours a business model in which intellectual property and research remain within the country while international customers pay to access AI capabilities through controlled services.
Alfredo Montufar-Helu, a Beijing-based managing director at consultancy Ankura, said the approach reflected a direction in which the market was already moving: retaining intellectual property and research and development domestically while selling access and services abroad.
This model could allow Chinese developers to generate revenue from international markets without transferring ownership of their most advanced technologies.
Companies such as DeepSeek, Moonshot and Z.AI already provide AI models to overseas customers through websites and cloud platforms. Under the proposed framework, such arrangements could remain broadly permissible, while certain overseas deployments and transfers of underlying technology would require government approval.
The approach could offer Beijing two advantages. First, it would help Chinese AI companies compete internationally by making their services available to foreign users. Second, it would reduce the risk of valuable technology, technical expertise and strategic capabilities passing into foreign hands.
The restrictions also reflect security concerns. Advanced AI systems can support sensitive data processing and potentially powerful cyber operations. Keeping their underlying technology under domestic control would give Chinese authorities greater oversight of how these capabilities are transferred and used.
How US Chip Restrictions Complicate China’s AI Expansion
The proposed rules come as Chinese AI developers face a difficult computing environment shaped by US export controls.
Advanced AI models require substantial computing power for training and deployment. Washington’s restrictions on exports of advanced Nvidia processors to China have encouraged some Chinese developers to seek computing capacity in overseas data centres, where access to more powerful chips may be available.
However, the new Chinese rules could complicate that strategy.
Researchers at major Chinese AI laboratories are seeking clarification on whether they can train or operate their most advanced models on foreign-owned servers located outside China, according to a person familiar with the discussions.
The uncertainty could affect companies that depend on international computing infrastructure to develop and deliver their products.
In a worst-case scenario, developers might need to maintain separate domestic and overseas versions of the same downloadable model, the source said. Such a requirement could increase operating costs, complicate software development and make international distribution more difficult.
The draft provides simplified approval procedures for Chinese companies building and operating their own computing facilities abroad. However, it does not explicitly offer the same treatment for transferring technology to foreign customers seeking to build equivalent facilities, according to legal analysis by Han Kun.
The result could be a more complicated international AI market in which companies must navigate restrictions imposed by both Beijing and Washington.
Which AI Technologies Could Face Export Restrictions?
The proposed rules establish several tests for determining which AI models would fall within the restrictions.
These include the computing required to train a model, measured against a threshold of (10^{26}) calculations; model size, measured at a threshold of two trillion parameters; and performance among the world’s top 10 systems in certain assessments of capability, use or efficiency.
Parameters are numerical values learned during training that help determine how an AI model processes information and generates outputs.
The proposed criteria could extend beyond the largest general-purpose AI systems. A smaller model that performs exceptionally well in a specialised task could potentially fall within the rules under the ranking-based test.
However, the draft does not specify which international rankings would determine eligibility, leaving companies uncertain about how regulators would apply the criteria.
Legal experts have also raised questions about how model size would be calculated for systems that activate only some of their components for each task. Han Kun has separately questioned whether a proposed provision concerning chip-design technology could extend beyond the most advanced products because it does not specify a minimum performance threshold.
Liu Jing, a senior partner at Beijing Dacheng Law Offices who participated in consultations with policymakers, said companies wanted greater clarity on the tests used to determine which models would be covered.
These uncertainties could influence investment decisions, international partnerships and the timing of product launches while companies wait for more detailed implementation guidance.
Why Foreign Investors Could Face Greater Scrutiny
The proposed controls could also affect corporate acquisitions, foreign investment and ownership arrangements involving Chinese AI technology.
Registration requirements that took effect on October 16 cover certain transfers that place Chinese technology under foreign ownership or control, including qualifying transactions already underway, according to an analysis by law firm Fangda.
The move follows Beijing’s earlier decision to order Meta to unwind its acquisition of Chinese-founded AI startup Manus, a deal valued at more than $2 billion.
That episode demonstrated the potential political and regulatory risks facing foreign companies seeking to acquire Chinese AI businesses.
The new framework could make similar transactions more complicated by requiring greater scrutiny of whether an acquisition or ownership change would transfer strategically important technology abroad.
For international investors, the implications extend beyond the purchase of AI companies. Access to models, computing infrastructure, research capabilities and technical expertise could also become relevant to regulatory decisions.
Companies may therefore need to assess whether their corporate structures and overseas partnerships comply with both Chinese export controls and the investment restrictions of other jurisdictions.
China and the United States Are Building Competing Technology Barriers
The emerging Chinese framework illustrates how technological competition between Washington and Beijing is increasingly operating in both directions.
The United States has restricted China’s access to advanced semiconductors, computing hardware and certain forms of investment. Beijing is now seeking to exercise greater control over the export of AI models, technical expertise and other valuable technological assets.
Yang Jingru, a partner at law firm Jincheng Tongda and Neal, described the approaches as opposing sides of the same strategic contest: Washington manages the inflow of computing power, hardware and capital, while Beijing is beginning to manage the outflow of models and technological assets.
The two countries are therefore seeking to protect different parts of the AI supply chain.
For Washington, restrictions on advanced chips aim to limit China’s access to technologies with significant economic and national security implications. For Beijing, export controls could preserve domestic advantages in AI development while preventing foreign competitors from freely acquiring Chinese capabilities.
This dynamic could encourage a more fragmented global technology market, with companies facing different requirements depending on where their models are developed, hosted, licensed and sold.
What Happens Next?
China’s proposed export controls remain subject to public consultation, and the final rules will determine how broadly the restrictions are applied. The November 15 deadline is an important next step, while the expected adoption by the end of the year could provide greater clarity for AI developers and investors.
The central challenge for Beijing will be balancing technological security with the commercial ambitions of its AI industry. Excessively restrictive rules could make it harder for Chinese companies to attract overseas customers and build international partnerships. More targeted controls, however, could allow them to expand their global presence while retaining ownership of strategically important technology.
For international businesses, the proposals signal that access to Chinese AI should not be confused with unrestricted rights to its underlying technology.
The broader significance is that competition between China and the United States is moving beyond control of semiconductor supplies toward control of the AI systems those chips help create. As both governments seek to protect their technological advantages, the ability to access, transfer and commercialise artificial intelligence could become another major front in the struggle for global technological leadership.
With information from Reuters.
