Beijing proposes adding 40 artificial intelligence, solar and biomedicine technologies to its export control catalogue, imposes an immediate registration requirement on exporters of the newly listed items, and gives the world until 15 November to comment on a regime that reaches further up the technology stack than any previous Chinese measure.
BEIJING, OCTOBER 19, 2026
China’s Ministry of Commerce and Ministry of Science and Technology published a draft revision of the national catalogue of technologies prohibited or restricted from export on 16 October, proposing to add 40 technologies across artificial intelligence, photovoltaic solar manufacturing and biomedicine. The draft was issued as Announcement 2026 No. 48 and is open for public comment until 15 November.
The announcement carries an immediate operative effect that is easy to miss alongside the draft itself. From 16 October, exporters of the newly listed technologies must register with the Ministry of Commerce before exporting. The registration requirement is not a licensing requirement and does not constitute approval or authorisation, but it establishes a reporting obligation on transactions that were unregulated the day before the announcement appeared. In the classification maintained by the Global Trade Alert monitoring project, the measure was recorded as three distinct interventions, one of which, an export-related non-tariff measure, was logged as already in force on the day of announcement and assessed as harmful to the trading interests of 164 jurisdictions.
The catalogue itself is not new. It was first published in 2001 and has been revised in 2008, 2020, 2023 and 2025. What is new is the depth of the proposed additions and the explicit reach into software, model weights and know-how rather than into physical goods.
What is being added
The 40 proposed additions split into 14 in the prohibited category, where export is not permitted at all, and 26 in the restricted category, where export requires a licence. By sector, 23 relate to artificial intelligence, 12 to photovoltaic solar manufacturing, and 5 to biomedicine.
On the artificial intelligence side, the prohibited list covers core technologies for large models trained on data classified as core data under China’s data security framework, AI agents whose principal application is in critical national information infrastructure, defence or core government affairs, technologies with significant offensive cyber capability, and high-performance AI detection and identification technologies.
The restricted list is broader and, for commercial actors, more consequential. It covers models trained on data classified as important data, and it establishes three alternative thresholds any one of which brings a model within scope: training compute of 10 to the power of 26 floating point operations, at least 2 trillion parameters, or capability assessed as globally advanced. It also covers other high-performance AI agents, design technologies for 2.5D and 3D AI chips and 3D flash memory, large computing clusters with particular attention to those exceeding 10,000 accelerator cards, AI-enabled cybersecurity technologies, and speech and voiceprint recognition together with Chinese-language translation technology.
The thresholds deserve attention because they are the first time a major jurisdiction has written quantitative model-capability criteria into an export control instrument in this form. A compute threshold of 10 to the 26 floating point operations and a parameter count of 2 trillion are both above the frontier of publicly disclosed Chinese models, which means the immediate binding effect is limited. The third criterion, globally advanced capability, has no published metric and is therefore the provision that will determine the regime’s actual reach.
The carve-outs matter as much as the listings
The draft contains exclusions that materially shape its commercial effect. Technologies already in the public domain through open-source release, published academic papers or similar channels are excluded from control. That exclusion is significant given the volume of Chinese AI research published openly and the prominence of Chinese open-weight model releases over the past two years.
More consequentially for international users, access by overseas parties to models deployed inside China, whether through application programming interfaces, web interfaces, software-as-a-service arrangements or lightweight on-device deployment, is not treated as a technology export. Localised deployment of a restricted technology overseas, by contrast, is an export and falls within the regime.
That distinction draws a clear line. A foreign company consuming a Chinese model as a cloud service is outside the regime. A foreign company taking model weights and running them on its own infrastructure is inside it. The practical effect is to push international use of Chinese AI toward hosted arrangements in which the Chinese party retains control of the artefact, which is a commercial architecture Chinese providers have reason to prefer in any event.
The draft also contemplates an automatic licensing mechanism under which certain restricted exports would be permitted essentially on filing. Candidate categories identified include open-source release, overseas deployment where the Chinese party retains control through its own overseas cloud infrastructure, and other forms of cross-border technical collaboration. The operational detail of automatic licensing is left to future rules.
Solar: controlling the equipment, not just the know-how
The photovoltaic additions are the most commercially immediate. Twelve technologies covering polysilicon production, wafer manufacture and photovoltaic cell production are proposed for control. Perovskite solar cell technology moves to the prohibited list. Heterojunction cell technology moves to the restricted list.
The feature of the solar provisions that trade practitioners have flagged is that the controls extend to core manufacturing equipment that embodies or supports the controlled technologies, not only to the technical know-how itself. Analysts at Baker McKenzie noted that this approach has not previously been applied to other categories in the catalogue and suggested it is aimed at overseas solar manufacturing facilities.
That is a meaningful extension. A technology export control that captures only documentation, training and technical assistance is difficult to enforce and relatively easy to work around. A control that captures the production line equipment reaches the physical bottleneck. Chinese equipment suppliers dominate global photovoltaic manufacturing tooling, and a company building a cell plant outside China without access to Chinese equipment faces a materially harder problem than one that merely lacks a Chinese technical partner.
The sectoral logic is visible. Perovskite and tandem architectures are the next generation of photovoltaic technology, and Chinese producers hold a substantial research and early commercialisation lead. Heterojunction is the current high-efficiency mainstream. Mature technologies remain uncontrolled. An executive in the Chinese solar industry, speaking anonymously to Global Times, characterised the approach as leaving mature manufacturing technologies open while safeguarding advanced ones.
Biomedicine and the licensing economy
The five biomedicine additions cover cell and gene therapy, small nucleic acid drugs, and pathogenic microorganism technologies. All pathogenic microorganism technologies are proposed for the prohibited category.
For cell and gene therapy and small nucleic acid drugs, the prohibition applies to reusable foundational technology platforms developed through long-term research and capable of supporting multiple products, and it applies only to transfers of intellectual property ownership. Licensing, contract research and development, and production services are explicitly excluded from the prohibition, though they remain subject to licensing where the underlying technology falls in the restricted category. Restricted items include more targeted technologies such as target discovery and validation, where novelty is defined as not publicly disclosed anywhere in the world.
This architecture is carefully drawn, and the reason is commercial. Chinese pharmaceutical innovation has become a significant source of licensed assets for the global industry. Overseas licensing deals for Chinese innovative drugs exceeded 130 billion US dollars in 2025 according to figures cited by Chinese state media, and roughly 30 percent of the global innovative drug development pipeline now originates in China. A control regime that impeded that licensing business would be self-defeating.
The draft therefore permits the flow that generates revenue, licensing without transfer of ownership, while restricting the flow that transfers durable capability, ownership of the underlying platform. The draft contemplates facilitation measures, including automatic licensing, for deals that do not transfer ownership, and for public health emergencies, humanitarian relief and cross-border research using overseas resources.
Beijing’s framing
A Ministry of Commerce spokesperson described the revision as a routine update conducted under existing law and said it does not target any specific country or region. The spokesperson added that China’s support for mutually beneficial international technology cooperation is unchanged.
Chinese commentators have reinforced that framing. Cui Fan, a professor at the University of International Business and Economics, said the catalogue regulates the order of the technology export market and protects national economic and technological interests, and noted that comparable controls exist in the United States, the European Union, Japan, the United Kingdom and South Korea. Zhao Qiuyan, a researcher at the China Machine Think Tank, described the approach as tiered according to end use, with core large models prohibited and important ones restricted.
The comparison to other jurisdictions is accurate as far as it goes. Technology export controls are a standard instrument, and the United States in particular has used them aggressively against Chinese semiconductor and AI capability. Reading this measure purely as retaliation would be too simple. But reading it as routine would also be a mistake. A revision that adds quantitative model thresholds, captures manufacturing equipment, and imposes an immediate registration obligation is a step change in the instrument’s reach.
The international reaction
The European Commission has signalled wariness. Reporting following the announcement indicated Commission concern about the scope of China’s export control plans affecting key technology components, in a period when Brussels and Beijing have been working through a parallel set of trade negotiations that include export control dialogue as an explicit workstream.
That overlap is awkward. The second round of EU-China Trade and Investment Consultations concluded in Beijing on 9 October with an agreement to strengthen dialogue on export controls and to explore smoother flows for compliant dual-use exports, including a Chinese indication that it would continue facilitating licences for rare earths and permanent magnets. One week later, Beijing published a draft that expands export control coverage into artificial intelligence and solar manufacturing.
The two actions are not formally inconsistent. The consultations addressed dual-use goods licensing; the catalogue addresses technology transfer. But the sequencing will complicate European assessments of whether the October understandings represent a durable easing or a tactical one.
Economic impact
The immediate measurable impact is on transaction cost rather than on trade volume. The registration requirement effective 16 October applies to exporters of newly listed technologies and creates an administrative step where none existed. For large Chinese technology firms with established compliance functions, that is manageable. For smaller firms, research institutions and joint ventures, it introduces uncertainty about whether a given collaboration falls within scope.
The larger impact is on investment planning. Any foreign company evaluating a technology partnership with a Chinese counterpart in AI, advanced photovoltaics or the listed biomedicine categories must now assume that the arrangement may require Chinese government licensing and that the licensing framework is in flux. That assumption changes the risk weighting on such deals regardless of whether a licence would in fact be refused.
For the solar industry specifically, the equipment provisions intersect with a wave of photovoltaic manufacturing investment outside China driven by industrial policy in the United States, the European Union and India. Projects that assumed access to Chinese production equipment and process technology now face a control regime over both. Projects already commissioned are unaffected. Projects in planning face a new variable.
For the artificial intelligence sector, the practical effect in the near term is modest because the quantitative thresholds sit above publicly disclosed Chinese model capability. The chip design provisions covering 2.5D and 3D AI chip architectures and 3D flash memory are more immediately binding, and they are notable as an assertion of Chinese capability worth protecting in a domain where the prevailing narrative has been one of Chinese disadvantage.
Implications for importers, exporters and supply chains
Companies with China-based research and development should audit what technology their Chinese entities have developed locally and whether moving that technology to affiliates elsewhere would now constitute a controlled export. Intra-group transfers are exports for these purposes, and multinational structures that routinely move technical know-how across borders are exposed in ways that arm’s-length trading relationships are not.
Companies deploying Chinese-developed AI models or agents outside China should determine whether their deployment architecture places them inside or outside the regime. The line between hosted access and localised deployment is the operative distinction, and it is drawn in the draft rather than in general principle. Firms running Chinese open-weight models on their own infrastructure should confirm whether the specific model release falls within the open-source exclusion.
Companies in photovoltaic manufacturing outside China should map their equipment and process technology dependencies against the listed categories and assess licensing exposure on projects not yet commissioned. Where alternatives exist, the cost differential should be reassessed against the regulatory risk.
Companies in pharmaceuticals and biotechnology should review in-licensing arrangements with Chinese partners. Deals structured as licences without ownership transfer appear to be outside the prohibition and candidates for facilitated treatment. Deals structured as acquisitions of underlying platform technology may not be.
And every affected company should consider whether to participate in the comment process. The window closes on 15 November. Chinese regulatory consultations have in the past produced material changes between draft and final text, and the provisions most likely to move are those where industry can demonstrate that the drafting captures transactions the regulator did not intend to reach.
What to watch
The first marker is the final text, expected after the comment period closes. Whether the globally advanced capability criterion is given an objective metric, whether the equipment provisions survive in their current form, and whether automatic licensing is defined operationally will determine the regime’s practical weight.
The second is enforcement posture. A registration requirement without published enforcement practice is an uncertain obligation. How the Ministry of Commerce handles the first cases, and whether registration becomes a de facto approval gate, will tell exporters more than the text does.
The third is reciprocal action. Export controls invite responses, and the question for the coming months is whether Brussels, Tokyo, Seoul and Washington treat this as a defensive measure within an accepted norm or as an escalation requiring answer. The answer will shape the environment in which every technology-intensive supply chain operates through 2027.
The catalogue as an instrument
It is worth being precise about what the catalogue is, because it is frequently conflated with China’s other export control instruments and the three operate differently.
The Export Control Law, in force since December 2020, governs dual-use items, military items and nuclear items, and is the authority under which China’s rare earth and gallium measures have been taken. It is primarily a goods regime with a technology component.
The unreliable entity list and the various countermeasure instruments are targeted at specific foreign firms and are explicitly retaliatory in design.
The Catalogue of Technologies Prohibited or Restricted from Export is older than both and sits in a different legal family. It derives from the foreign trade framework and regulates the export of technology as such, meaning know-how, designs, processes and now model artefacts, independent of whether any physical good crosses a border. Its 2023 revision attracted international attention chiefly for adding technologies relevant to rare earth processing and to lithium battery cathode production.
Understanding the distinction matters for compliance. A company that has mapped its exposure to the Export Control Law has not thereby mapped its exposure to the catalogue, because the catalogue reaches transactions that involve no shipment at all. A training session, a design handover, a secondment of engineers, or the transfer of a model checkpoint can each be a technology export under this instrument.
The numbers behind Beijing’s confidence
Chinese commentary on the revision has leaned on a set of figures intended to establish that China now holds technology worth controlling. Chinese entities account for roughly 60 percent of global artificial intelligence patents according to World Intellectual Property Organization data cited in Chinese state media. Chinese innovative drug licensing to foreign partners exceeded 130 billion US dollars in value during 2025. Approximately 30 percent of the global innovative drug pipeline in development originates with Chinese organisations.
Patent counts are a weak proxy for capability and should be treated cautiously, since filing volume reflects incentive structures as much as inventive output. The pharmaceutical figures are more robust, because licensing deals are arm’s-length transactions in which sophisticated counterparties pay for assets they have evaluated.
The underlying point stands regardless of the specific metrics. For two decades Chinese technology policy was organised around acquiring capability from abroad. The catalogue revision is an artefact of a different posture, one in which the governing concern is retention rather than acquisition. That shift has implications for every multinational that structured its China strategy around technology transfer in, and it will take time for those structures to adjust.
A structural read
Set against the pattern of the past three years, this measure fits a recognisable logic. Jurisdictions that hold a position in a critical technology are moving to convert that position into leverage, and the instrument of choice is the export control rather than the tariff.
Tariffs tax trade. Export controls stop it. A tariff is a price signal that market participants can absorb, arbitrage or pass through. An export control on a technology with no substitute is a binary constraint, and the party holding the control holds a different kind of power from the party holding a tariff schedule.
The consequence for global supply chains is a shift in the nature of the risk firms must manage. Tariff risk is quantifiable, insurable in effect through pricing, and reversible. Control risk is none of those things. A company that loses access to a controlled technology cannot hedge the exposure financially; it can only redesign around the dependency, which takes years.
That asymmetry is why the registration requirement effective 16 October deserves more attention than the headline list of 40 technologies. The list is a draft subject to change. The registration obligation is in force now, and it establishes the administrative machinery through which a future tightening would operate.