On October 16, China’s Ministry of Commerce (MOFCOM) and Ministry of Science and Technology (MOST) issued Announcement No. 48, proposing revisions to China’s Catalogue of Technologies Prohibited or Restricted from Export, including 23 AI-related technologies.
At the same time, the authorities introduced an export registration requirement for the technologies listed in the announcement, effective immediately. This requirement also applies to exports that have already commenced but have not yet been completed.
Importantly, the registration requirement is already in effect, while the proposed additions to the export control catalogue remain subject to public consultation.
Pursuant to the Foreign Trade Law of the People’s Republic of China and the Regulations of the People’s Republic of China on the Administration of Technology Import and Export, the Ministry of Commerce and the Ministry of Science and Technology propose to revise the Catalogue of Technologies Prohibited or Restricted from Export to include certain technologies in artificial intelligence, biomedicine, and photovoltaics.
With the necessary approval, the proposed revisions are hereby released for public consultation. At the same time, a pre-export registration requirement will be introduced for the relevant technologies.
The public consultation covers 40 technologies across three sectors: artificial intelligence, biomedicine, and photovoltaics.
Artificial Intelligence: The proposed additions cover 23 technologies related to large AI models, AI agents, computing clusters, and information processing. Technologies that have already entered the public domain through open-source releases, academic publications, or other means will not be subject to these controls.
Biomedicine: The proposed additions cover five technologies related to cell and gene therapies, small nucleic acid drugs, and pathogenic microorganisms. Facilitation measures, including automatic licensing, will apply to overseas licensing arrangements that do not involve a transfer of ownership.
Photovoltaics: The proposed additions cover 12 technologies related to polysilicon manufacturing, silicon wafer production, and solar cell manufacturing. The scope of technology controls will also extend to key and core equipment incorporating the relevant technologies.
The specific technologies proposed for inclusion, together with explanations of the relevant policy measures, are provided in Annex 1.
The public may submit comments between October 16 and November 15, 2026, regarding the proposed scope of technologies, technical thresholds, control measures, and facilitation arrangements through the following channels:
Email: jinxianmulu@mofcom.gov.cn
Fax: +86-10-65197565
Postal mail: Bureau of Industry, Security, Import and Export Control, Ministry of Commerce, No. 2 Dong Chang’an Avenue, Beijing 100731, China.
Please indicate “Public Consultation on the Catalogue of Technologies Prohibited or Restricted from Export” in the email subject line, on the cover page of any fax, or on the envelope.
Submissions should include the name of a contact person, the organization they represent, and their contact details. Submissions made on behalf of an organization must bear its official seal.
(1) Effective Period and Scope
From the date of publication of this announcement until the revised Catalogue of Technologies Prohibited or Restricted from Export is formally issued and takes effect, technology exporters must register the export of any technology listed in the annexes to this announcement before proceeding with the export.
Registration must be completed through the Technology Trade Management Information System on MOFCOM’s Unified Business Platform:
https://ecomp.mofcom.gov.cn
The registration requirement also applies to technology exports that have already commenced but have not yet been completed.
(2) Transactions Subject to Registration
The following types of technology export transactions are subject to registration:
Transfers of technology ownership to overseas entities, including transfers of intellectual property rights.
Licensing of domestically held technologies for use overseas, including intellectual property licensing arrangements.
Changes in the ownership or actual control of domestically held technologies whereby the owner or actual controller becomes an overseas entity.
Other circumstances in which MOFCOM explicitly requires registration through notices or other official communications.
(3) Information Required for Registration
The information to be registered mainly includes:
Basic information about the technology exporter and the technology recipient.
The destination country or region.
Details of the technology proposed for export and its intended use.
The estimated value of the technology export.
The specific information requirements are subject to those set out in the Technology Trade Management Information System on MOFCOM’s Unified Business Platform, as detailed in Annex 2.
Where multiple technologies are intended to be exported to the same recipient as part of a single transaction, they may be included in one registration submission. In other circumstances, separate registrations should generally be submitted.
Annex 1: Proposed Scope of Artificial Intelligence, Biomedical, and Photovoltaic Technologies to Be Included in the Catalogue of Technologies Prohibited or Restricted from Export (Draft for Public Consultation).
Annex 2: Technology Export Information Registration Form.
Ministry of Commerce of the People’s Republic of China Ministry of Science and Technology of the People’s Republic of China
October 16, 2026
Serial number; industry sector; code; technology name; control points.
Internet and related services. High-performance detection technology: Technologies for manufacturing authentication systems and equipment specifically or mainly used for preventing and handling security risks in cross-border transmission of China’s network data, where the detection accuracy and recall rate for deeply synthesized video, audio, images, or multimodal information are 95%.
Information confrontation technology: 1. AI autonomous-driven malicious code generation and mutation technologies that have any of the following characteristics and may be used to attack highly sensitive systems such as national critical information infrastructure, national defense and military industry, and core government affairs, or to bypass national security defense systems: (1) an attack success rate of 60% against multi-engine malware detectors;
(2) an attack success rate of 85% against a single mainstream antivirus software;
(3) an attack success rate of 90% against machine learning models.
2. Adaptive evolution in adversarial environments and runtime dynamic behavior modification technologies that have any of the following characteristics and may be used to attack highly sensitive systems such as national critical information infrastructure, national defense and military industry, and core government affairs, or to bypass national security defense systems: (1) capable of detecting virtualization sandboxes, debuggers, and emulators, identifying the type and version of security products, or achieving sandbox escape;
(2) under the premise of maintaining complete malicious functionality, automatically rewriting 50% of code structure or control flow upon each execution.
3. Automated antivirus-evasion variant batch generation and distributed large-scale deployment technologies that have any of the following characteristics and may be used to attack highly sensitive systems such as national critical information infrastructure, national defense and military industry, and core government affairs, or to bypass national security defense systems: (1) generating 20 functionally equivalent but structurally different variants in a single run, or having an end-to-end automation rate of 80%;
(2) based on a single baseline payload, generating more than 10 variants every 24 hours with syntax tree divergence of 30%;
(3) sampled variants showing 90% consistency in behavioral profiles in sandboxes.
Software and information technology services. AI large model technology: Large models that provide services after being trained using China’s core data.
Note: The scope of “core data” is determined by the Data Security Law of the People’s Republic of China and other relevant laws and regulations.
Software and information technology services. AI agent technology: Agents specifically or mainly used in highly sensitive application scenarios such as national critical information infrastructure, national defense and military industry, and core government affairs, and reaching a high-performance level.
Note: “High performance” means meeting the control indicators for restricted-category AI agent technology under 266505X.
Serial number; industry sector; code; technology name; control points.
Internet and related services. High-performance detection technology: (Added on the basis of the existing seven items under this category of technology) 8. Technologies for manufacturing detection systems and equipment for deeply synthesized video, audio, images, or multimodal information, with accuracy and recall rate of 98%.
9. Video and image information steganography and extraction technologies that simultaneously meet the following characteristics: (1) high robustness, with an error code rate of 0.5% or recall rate of 99.9%;
(2) low false alarm rate, with false positive rate (FPR) of 0.1%;
(3) high imperceptibility, with structural similarity index (SSIM) of 0.999, or peak signal-to-noise ratio (PSNR) of 41 dB, or learned perceptual image patch similarity (LPIPS) of 0.005.
Information confrontation technology: (Added on the basis of the existing five items under this category of technology) 6. AI-assisted vulnerability discovery technologies with any of the following characteristics: (1) in authoritative security benchmark tests, a vulnerability reproduction or exploitability determination score of 75%;
(2) for known publicly disclosed vulnerabilities, an end-to-end success rate for independently completing the process from vulnerability analysis to construction of a complete exploit chain: 50% success rate for mainstream operating systems or browsers, or 40% success rate for exploit construction at the arbitrary-code-execution level;
(3) vulnerability discovery and exploitation efficiency achieving an improvement of 10 times or more compared with the benchmark implementation by senior security experts, or achieving a performance improvement of 10 times or more in the number of exploitable vulnerabilities discovered in standard vulnerability database security benchmark tests;
(4) capable of autonomously chaining two or more low-risk or medium-risk software defects into a complete multi-stage attack chain and generating runnable proof-of-concept code;
(5) under a conventional testing framework, capable of autonomously discovering in one pass more than 5 previously undisclosed high-risk security defects, or more than 50 previously undisclosed security defects, in a major open-source project or mainstream operating system;
(6) reaching the highest level in mainstream domestic and international exploit capability grading systems.
Software and information technology services. Information processing technology: (Added on the basis of the existing 18 items under this category of technology) 19. Voiceprint recognition technologies with any of the following characteristics, including core voiceprint algorithm technology, anti-attack and anti-forgery technology, voiceprint recognition system technology, voiceprint data processing and protection technology, etc.: (1) under conditions of false positive rate (FPR) of 0.01%, enrollment speech duration of 30 seconds, test speech duration of 10 seconds, and speech signal-to-noise ratio of 20 dB, a true positive rate (TPR) of 99.1%;
(2) single voiceprint end-to-end processing latency of 300 milliseconds, and throughput of 100 requests/second;
(3) overall detection rate of 98% for common fraud attacks such as recording replay, speech synthesis, and voice conversion, or false acceptance rate of 1%;
(4) possessing specialized optimization capabilities for voiceprint recognition of minority languages or specific regional dialects, and recognition accuracy of 90% under standard test conditions.
20. Translation technologies with any of the following characteristics, specifically or mainly used for Chinese and minority languages: (1) an average mean opinion score (MOS) of 4.8 obtained through human evaluation on standard test sets;
(2) where translation semantic fidelity is 95%, a BLEU similarity score with standard answers of 35 for general-scenario translation, or 30 for professional-domain translation.
Software and information technology services. AI large model technology: 1. Large models that provide services after being trained using China’s important data.
2. Core technologies of large models with any of the following characteristics, including weight parameters, key checkpoints, underlying source code, architecture design, and technologies such as training optimization and inference acceleration: (1) computational volume of 1026 floating-point operations;
(2) parameter count of 000 billion;
(3) capabilities reaching the world-leading level.
Notes: Where only a trace amount (0.1%) of important data is used for training, the important data used cannot substantially improve the model’s core technology R&D capability in that data field, security measures such as desensitization and cleaning are adopted, the important data will not be reproduced in the output stage, or the relevant important data can be legally and publicly obtained, it is not included in the scope of administration of item 1. The scope of “important data” is determined by the Data Security Law of the People’s Republic of China and other relevant laws and regulations.
“Capabilities reaching the world-leading level” means that, at the time of the export activity, the model ranks among the global top 10 in public evaluations by mainstream domestic and international evaluation institutions, large-scale industry platforms, etc., in terms of comprehensive performance of large models, specialized business capabilities, global call volume, safety performance, training efficiency, etc.
Software and information technology services. AI agent technology: 1. Autonomous planning and task decomposition technology, capable of automatically decomposing natural-language instructions into a chain of 20 executable subtasks, with subtask decomposition accuracy of 80% and average subtask success rate of 90%.
2. Multi-agent collaborative architecture technology, supporting 60 concurrent collaborative agents, with overall task completion time, under the same effect, being two times the minimum completion time of a single agent, and the consistency rate between the final multi-agent conclusion and the standard answer being 85%.
3. Tool calling and system operation technology, supporting integration of 200 third-party tools, with tool selection accuracy of 90%, and in scenarios requiring three or more tools to collaborate to complete a task, average end-to-end task success rate of 90%.
4. AI game decision-making technology with all of the following characteristics: (1) single-task decision planning depth of 20 steps or 10 concurrent game agents;
(2) possessing strategy self-adaptation capability, complex dynamic game decision latency of 200 milliseconds and 10 control units;
(3) exceeding 90% of humans on standard game benchmarks.
Software and information technology services. AI chip design technology: 1. AI processor chip design technologies with any of the following characteristics, including chip architecture, self-developed instruction sets, intelligent computing intellectual property cores, supporting foundational software stacks, etc.: (1) requiring use of 14 nm or below manufacturing process;
(2) possessing floating-point or integer computing capability;
(3) adopting 2.5D or 3D advanced packaging integration technology;
(4) possessing high-speed interconnection capability;
(5) new architectures based on non-von Neumann architectures such as compute-in-memory and neuromorphic computing;
(6) applicable to specific scenarios such as autonomous driving, computer vision, and intelligent speech.
2. Memory chip design technology: (1) 3D flash memory chip design technologies with any of the following characteristics: 1) storage density of 8 gigabits per square millimeter (Gb/mm2); 2) I/O rate of 1.5 gigabits per second (Gbps); 3) adoption of bonding technology. (2) Memory chip design technologies with any of the following characteristics: 1) requiring use of 18 nm or below memory manufacturing process; 2) adopting a vertical-channel transistor architecture; 3) adopting bonding technology; 4) implementing multilayer stacking.
Software and information technology services. AI computing power cluster technology: 1. Inter-chip or inter-node high-speed interconnection technology for large-scale intelligent computing clusters, with bidirectional bandwidth of 800 gigabytes per second (GB/s).
2. Distributed training fault-tolerance technologies with any of the following characteristics, used in large-scale intelligent computing clusters to ensure stable operation of distributed training: (1) average mean time to recovery (MTTR) of the checkpoint recovery mechanism of 120 seconds;
(2) supporting automatic fault sensing and seamless task migration, with visible interruption time caused by faults to training tasks of 30 seconds;
(3) under a typical cluster scale, the effective computing power waste rate of graphics processing units (GPUs) caused by network faults is 5%.
3. Efficient communication scheduling technologies with any of the following characteristics, used in large-scale intelligent computing clusters to optimize cluster communication efficiency: (1) linear scalability of 98%;
(2) possessing dynamic topology-aware routing and cross-multiple-plane load balancing capabilities, with end-to-end communication latency of 1 microsecond;
(3) supporting deep overlap of communication and computation, with communication hiding rate of 90%.
4. Heterogeneous-cluster efficient inference and scheduling technologies with any of the following characteristics, supporting unified scheduling and collaborative inference of AI chips from two or more different vendors in large-scale intelligent computing clusters, including heterogeneous computing power unified scheduling frameworks, mixed-precision collaborative training technologies, heterogeneous inference optimization technologies, heterogeneous-cluster intelligent operation and maintenance technologies, etc.: (1) supporting unified adaptation and hybrid scheduling of two or more high-speed communication protocols, with cross-chip communication bandwidth utilization of 90%, or cross-node end-side communication latency of 5 microseconds;
(2) supporting unified management of computing nodes across more than two different architectures, with throughput performance of heterogeneous inference clusters improved by 30% compared with a single-chip cluster solution, or average utilization of AI accelerator cards in heterogeneous clusters improved by 65%, or inference cost reduced by 20% compared with a single-chip solution, or cluster average fault recovery time of 10 minutes and cluster availability of 99.9%;
(3) supporting unified computing power scheduling and task allocation across two or more geographically distributed intelligent computing centers, with cross-region scheduling training efficiency being 80% of single-cluster efficiency, or supporting real-time load balancing and task migration for cross-region heterogeneous computing power resources.
Note: “Large-scale intelligent computing cluster” refers to a computing entity with more than 10,000 AI accelerator cards in total.
Serial number; industry sector; code; technology name; control points.
Pharmaceutical manufacturing. Original cell and gene therapy technology: Cell and gene therapy technology process platforms with independent intellectual property rights.
Notes: Only outbound transfer of ownership is prohibited; outbound licensing, use of relevant technology processes to provide outsourced R&D and production services, and other circumstances that do not transfer ownership outbound are not within the scope of administration. For specific drug varieties at different R&D and production stages, relevant technology processes necessary for R&D, testing, production, quality control, marketing, compliance, etc. in the course of outbound cooperation or business transactions (BD) are not within the scope of administration.
“Technology process platform” refers to a set of reusable underlying common technologies formed through long-term R&D accumulation and capable of supporting the development of multiple drugs. Development technologies and processes used for a single drug variety are not within this scope.
Pharmaceutical manufacturing. Original small nucleic acid drug technology: Small interfering ribonucleic acid (siRNA) drug technology process platforms with independent intellectual property rights.
Notes: Only outbound transfer of ownership is prohibited; outbound licensing, use of relevant technology processes to provide outsourced R&D and production services, and other circumstances that do not transfer ownership outbound are not within the scope of administration. For specific drug varieties at different R&D and production stages, relevant technology processes necessary for R&D, testing, production, quality control, marketing, compliance, etc. in the course of outbound cooperation or business transactions (BD) are not within the scope of administration.
“Technology process platform” refers to a set of reusable underlying common technologies formed through long-term R&D accumulation and capable of supporting the development of multiple drugs. Development technologies and processes used for a single drug variety are not within this scope.
Research and experimental development. Pathogenic microorganism technology: Artificial genetic modification and directed passage gain technologies for pathogenic microorganisms, with any of the following characteristics: (1) function-enhancement technologies for known or potential pathogenic microorganisms;
(2) passage culture systems and processes used for large-scale proliferation of pathogenic microorganisms, directed accelerated evolution, and stable maintenance of specific virulence characteristics;
(3) artificial controlled passage induction technologies simulating natural evolutionary processes.
Serial number; industry sector; code; technology name; control points.
Pharmaceutical manufacturing. Original cell and gene therapy technology: Completely novel target discovery and validation technologies for cell and gene therapy with independent intellectual property rights.
Notes: “Completely novel target” refers to a drug target that has not been publicly disclosed worldwide through patents, academic literature, public databases, industry technical materials, academic conferences, or other channels.
“Completely novel target discovery and validation technology” refers to technology used to screen, discover, and confirm completely novel targets and to identify drug action sites with druggability potential.
Pharmaceutical manufacturing. Original small nucleic acid drug technology: Completely novel target discovery and validation technologies for small interfering ribonucleic acid (siRNA) with independent intellectual property rights.
Notes: “Completely novel target” refers to a drug target that has not been publicly disclosed worldwide through patents, academic literature, public databases, industry technical materials, academic conferences, or other channels.
“Completely novel target discovery and validation technology” refers to technology used to screen, discover, and confirm completely novel targets and to identify drug action sites with druggability potential.
Serial number; industry sector; code; technology name; control points.
Non-metallic mineral products industry. Polysilicon preparation technology: 1. Cold hydrogenation fluidized-bed process with single-line trichlorosilane production capacity of 250,000 tons/year, and technologies for manufacturing its equipment.
2. Silane fluidized-bed process with single-line capacity of 20,000 tons/year, and technologies for manufacturing its equipment.
Non-metallic mineral products industry. Silicon wafer preparation technology: Silicon wafer preparation technology with thickness of 70 micrometers, and technologies for manufacturing its equipment.
Note: “Silicon wafer preparation technology” includes program code, underlying algorithm models, equipment drawings, etc.
Electrical machinery and equipment manufacturing. Photovoltaic cell preparation technology: 1. Edge passivation photovoltaic-specific technology for crystalline silicon cell preparation, and technologies for manufacturing its equipment, simultaneously meeting the following conditions: (1) usable for full, half, or multi-split solar silicon wafers, and calculated by the size of a full silicon wafer, each side length is between 182 and 230 millimeters;
(2) silicon wafer thickness of 140 micrometers.
2. Perovskite and tandem cell module production technologies, and technologies for manufacturing their equipment, including: (1) perovskite-crystalline silicon tandem cell technology, with mass-produced silicon wafer size of 182 millimeters and efficiency of 30%;
(2) perovskite-specific slot-die coating technology, with processing width of 1.1 meters;
(3) large-area single-junction perovskite cell module technology, with single-junction perovskite cell module area of 1 square meter and efficiency of 18%;
(4) large-area tandem perovskite cell module technology, with module area of 1 square meter and efficiency of 28%.
Serial number; industry sector; code; technology name; control points.
Non-metallic mineral products industry. Polysilicon preparation technology: 1. Polysilicon engineering technology with single-line capacity of 30,000 tons/year.
2. Reduction process with single reduction furnace polysilicon capacity of 800 tons/year, and technologies for manufacturing its equipment.
Note: “Engineering technology” includes equipment engineering design, plant construction engineering design, process packages, etc.
Non-metallic mineral products industry. Silicon wafer preparation technology: 1. Technologies for pulling photovoltaic monocrystalline silicon rods, and technologies for manufacturing their equipment, including: (1) monocrystalline furnace design application schemes with furnace body chamber inner diameter of 1600 millimeters;
(2) monocrystalline furnace thermal field design application schemes with size of 35 inches;
(3) monocrystalline furnace design application schemes with pulling speed of 1.2 millimeters/minute;
(4) monocrystalline furnace design application schemes with constant-diameter power consumption of 40 kilowatts.
2. Machining technologies for photovoltaic monocrystalline silicon rods, and technologies for manufacturing their equipment, including: (1) half-rod cutting and grinding preparation processing technology;
(2) machining technology usable for manufacturing silicon rods with length of 720 millimeters.
3. Ultra-thin large-size photovoltaic silicon wafer preparation technologies, and technologies for manufacturing their equipment, including: (1) preparation technologies usable for manufacturing rectangular silicon wafers with any side length of 182 millimeters;
(2) technologies usable for preparing silicon wafers with thickness of 130 micrometers or below: 1) tungsten-wire slicing technology of 28 micrometers or below; 2) slicing equipment design application schemes with spindle speed of 6500 revolutions/minute, or main roller diameter of 120 millimeters, or transverse axial distance of main rollers of 370 millimeters.
Electrical machinery and equipment manufacturing. Photovoltaic cell preparation technology: 1. Preparation technologies for high-efficiency tunnel oxide passivated contact (TOPCon) and high-efficiency back contact (XBC) crystalline silicon cells, and technologies for manufacturing their equipment, including: (1) ultra-thin tunnel oxide layer and doped polysilicon selective passivated contact layer low-pressure chemical vapor deposition (LPCVD) and plasma-enhanced chemical vapor deposition (PECVD) photovoltaic-specific coating technologies simultaneously meeting the following conditions: 1) usable for full, half, or multi-split solar silicon wafers, and calculated by the size of a full silicon wafer, each side length is between 182 and 230 millimeters; 2) silicon wafer thickness of 140 micrometers; 3) calculated by a full G12 silicon wafer, LPCVD equipment capacity of 6000 wafers/hour and PECVD equipment capacity of 6000 wafers/hour. (2) Aluminum oxide surface passivation layer atomic layer deposition (ALD) photovoltaic-specific coating technologies simultaneously meeting the following conditions: 1) calculated by the size of a full silicon wafer, each side length is between 182 and 230 millimeters; 2) silicon wafer thickness of 140 micrometers; 3) calculated by a full G12 silicon wafer, capacity of 12000 wafers/hour. (3) High-temperature boron diffusion photovoltaic-specific technologies simultaneously meeting the following conditions: 1) calculated by the size of a full silicon wafer, each side length is between 182 and 230 millimeters; 2) silicon wafer thickness of 140 micrometers; 3) calculated by a full G12 silicon wafer, capacity of 4500 wafers/hour. (4) TOPCon cell copper electroplating electrode preparation technologies simultaneously meeting the following conditions: 1) usable for full, half, or multi-split solar silicon wafers, and calculated by the size of a full silicon wafer, each side length is between 182 and 230 millimeters; 2) silicon wafer thickness of 140 micrometers. (5) Low-temperature amorphous or microcrystalline silicon deposition and PECVD photovoltaic-specific coating technologies simultaneously meeting the following conditions: 1) usable for full, half, or multi-split solar silicon wafers, and calculated by the size of a full silicon wafer, each side length is between 182 and 230 millimeters; 2) silicon wafer thickness of 140 micrometers. (6) XBC cell electroplating and chemical coating technologies simultaneously meeting the following conditions: 1) used for full, half, or multi-split solar silicon wafers, and calculated by the size of a full silicon wafer, each side length is between 182 and 230 millimeters; 2) silicon wafer thickness of 140 micrometers.
2. Heterojunction (HJT) crystalline silicon cell preparation technologies, and technologies for manufacturing their equipment, including: (1) heterojunction cell stacked transparent conductive film deposition technologies with any of the following characteristics, usable for processing silicon wafers with thickness of 110 micrometers: 1) indium content in transparent conductive film of 50%; 2) calculated by G12 silicon wafers, capacity of 14000 half-wafers/hour. (2) heterojunction cell crystalline silicon film deposition technologies with any of the following characteristics, usable for processing silicon wafers with thickness of 110 micrometers: 1) used for high-efficiency N-type heterojunction cells, with film crystallization rate of 40%, or calculated by G12 silicon wafers, capacity of 14000 half-wafers/hour; 2) used for P-type heterojunction cells, with maximum conversion efficiency of 23% under AM1.5 test conditions, or maximum conversion efficiency of 20.5% under AM0 test conditions. (3) Low-silver silver-coated copper heterojunction solar cell preparation technology, with silver content of 30%. (4) Copper electroplating electrode heterojunction cell preparation technology.
Note: In this category of technologies, “high efficiency” means average cell conversion efficiency of 25% under AM1.5 test conditions, or average mass-produced module conversion efficiency of 22. 3%.
Technologies that have entered the public domain through open source, publication of academic papers, or other means are not included in the scope of administration.
For technology exports involving open source of this category of technologies, as well as Chinese enterprises’ self-built and self-operated computing power centers overseas, overseas chip tape-out manufacturing, joint R&D using overseas resources, participation in international standard setting, and other circumstances, facilitation measures such as automatic licensing shall apply.
For technology exports involving public health emergencies, humanitarian assistance, joint R&D using overseas resources, outbound licensing, and other circumstances, facilitation measures such as automatic licensing shall apply.
The scope of technology administration also includes key and core equipment carrying the relevant technologies, specifically including trichlorosilane cold hydrogenation fluidized beds, silane fluidized beds, reduction furnaces, photovoltaic silicon wafer slicing equipment, monocrystalline furnaces applied to photovoltaic crystal pulling, machining equipment for photovoltaic silicon wafer preparation such as truncating and squaring, photovoltaic silicon wafer slicing equipment, edge passivation equipment (EPD) applied to half-wafer or multi-split processes, tunnel oxide doped polysilicon selective passivated contact layer deposition equipment (including both PECVD and LPCVD equipment), atomic layer deposition (ALD) equipment, high-temperature boron diffusion furnaces, copper electroplating equipment applied to TOPCon cells, crystalline silicon film deposition (PECVD) equipment applied to HBC and HTBC cells, copper electroplating equipment applied to XBC cells, transparent conductive film deposition equipment (PVD), amorphous or microcrystalline silicon film deposition equipment (PECVD), heterojunction cell copper electroplating equipment, and slot-die coating equipment.

On October 16, 2026, the Ministry of Commerce (MOFCOM) and the Ministry of Science and Technology (MOST) issued Announcement No. 48 of 2026, launching a public consultation on proposed revisions to the Catalogue of Technologies Prohibited or Restricted from Export (the “Catalogue”) and introducing temporary pre-export registration requirements for the relevant technologies.
A MOFCOM spokesperson responded to questions from the media regarding the announcement.
A: Under Article 20 of the Foreign Trade Law of the People’s Republic of China, the competent foreign trade department of the State Council, together with other relevant departments, is responsible for formulating, adjusting, and publishing the catalogue of technologies prohibited or restricted from export.
Since 2001, MOFCOM and MOST have revised the Catalogue on multiple occasions, adding some technologies, removing others, introducing controls where necessary, and relaxing restrictions where appropriate. This has become an established regulatory practice.
The current revision is likewise part of China’s routine technology trade management, conducted in accordance with applicable laws and regulations. It is not directed at any particular country, region, or entity.
In recent years, technologies related to artificial intelligence, biomedicine, and photovoltaics have increasingly been applied in sensitive areas. Major economies have introduced export controls of varying degrees in these sectors.
Drawing on internationally accepted practices, China is implementing necessary regulatory measures in accordance with the law. The objective is to prevent the risks of technology becoming uncontrollable or spreading inappropriately, while safeguarding normal international technological exchanges and cooperation, and ensuring that technological development remains beneficial, safe, controllable, and serves humanity.
A: The public consultation covers 40 key technologies across three sectors: artificial intelligence, biomedicine, and photovoltaics.
Fourteen technologies are proposed to be prohibited from export, including AI-driven malicious code generation and mutation technologies, proprietary technology platforms and processes for cell and gene therapies, and production technologies for perovskite and tandem solar cell modules.
Twenty-six technologies are proposed to be restricted from export, including AI-assisted vulnerability discovery technologies, proprietary technologies for discovering and validating novel targets for general-purpose cell and gene therapies, and polysilicon engineering technologies with a production capacity of at least 30,000 tonnes per year per production line.
The announcement also specifies circumstances in which facilitation measures may apply. Further details are available in the original announcement.
A: The Chinese government has consistently supported enterprises in conducting mutually beneficial international business operations and technological cooperation in accordance with laws and regulations. This policy position remains unchanged.
Before launching the public consultation, MOFCOM and MOST, together with the relevant industry authorities, organized extensive assessments involving research institutions and industry associations. They also consulted major enterprises and carefully considered the current state and development needs of industries both in China and overseas.
This is reflected in three main aspects.
First, the scope of controls is precise and limited. Most of the technologies proposed for inclusion are defined by specific technical parameters. Products manufactured using these technologies are not subject to the controls. This provides enterprises with clear and narrowly defined compliance boundaries and facilitates implementation.
Second, facilitation measures will be available. The proposed controls identify various export scenarios eligible for simplified procedures, reflecting a differentiated and category-based regulatory approach. This is intended to improve policy transparency and predictability while ensuring that technology exports remain safe, compliant, and orderly.
Third, the Catalogue will be dynamically adjusted. MOFCOM and MOST will continue to adjust the scope of the Catalogue in accordance with the law and technological developments, removing technologies that are no longer advanced or have become obsolete.
The current public consultation is intended to gather views and recommendations from a broader range of stakeholders.
A: Global AI innovation is entering a period of unprecedented activity, bringing both opportunities and challenges.
The proposal to add 23 key technologies related to large AI models, AI agents, computing clusters, and information processing is based primarily on two considerations.
First, ensuring that AI technologies remain safe and controllable.
The inherent and associated risks of AI have attracted significant attention worldwide.
Some of the technologies proposed for control, such as AI-driven autonomous malicious code generation and mutation technologies, could be used to attack sensitive systems, including national critical information infrastructure, defense and military systems, and core government systems, or to circumvent national security defenses.
If acquired by non-state actors or malicious entities, these technologies could pose risks to national security and the shared security of humanity.
The Chinese government’s purpose in introducing necessary controls is to establish safeguards against misuse and malicious applications, ensure that AI remains under human control, and advance international AI governance.
Second, encouraging open source development, openness, cooperation, and technology sharing.
The Chinese government is committed to making AI an important driver of shared prosperity and common security.
To support the development of an open-source AI innovation ecosystem, facilitation measures, including automatic licensing, will apply to several categories of technology exports, including:
Open-source releases of AI technologies.
Chinese companies establishing and operating their own computing centers overseas.
Overseas semiconductor tape-out and manufacturing activities.
Joint research and development using overseas resources.
Participation in international standards development.
These measures are intended to allow exports to proceed with minimal delay, essentially on a “register and export” basis.
Technologies that have already entered the public domain through open-source releases, academic publications, or other means are proposed to be excluded from export controls, creating a favorable environment for international AI exchanges and cooperation.
It should be particularly emphasized that China advocates AI dialogue and cooperation with all countries on the basis of mutual respect, equality, and mutual benefit, in order to jointly address the risks of AI proliferation and loss of control.
China supports Global South countries in strengthening their AI capabilities, bridging the digital divide, and promoting sustainable development.
Additional facilitation and support will be provided for technology exports required for international cooperation among members of the World Artificial Intelligence Cooperation Organization, including capacity building, technology sharing, and joint development of rules.
A: Chinese AI models and applications are expanding internationally through a wide range of business models.
The key factor in determining whether a particular form of overseas expansion falls within the scope of technology export controls is whether it involves the transfer of restricted technologies from within China to overseas destinations.
The scope of AI technologies proposed for control is precise and limited. Once the relevant measures formally take effect, the vast majority of AI technologies will remain freely exportable.
For AI technologies classified as restricted:
Overseas local deployment falls within the scope of technology export controls. However, where deployment takes place on a company’s own overseas cloud infrastructure or under other relatively controllable circumstances, facilitation measures such as automatic licensing may apply.
By contrast, where enterprises use restricted large AI models to provide API access, interactions through apps or websites, SaaS applications, or lightweight deployment on local end-user devices, these activities do not fall within the scope of large AI model technology export controls.
A: The existing Catalogue already includes certain biomedical technologies. The current proposal would add five technologies related to pathogenic microorganisms, cell and gene therapies, and small nucleic acid drugs.
There are three main considerations.
First, protecting human health and safety.
Technologies involving pathogenic microorganisms are directly related to public health and national biosecurity.
Prohibiting the export of these technologies is an institutional measure intended to safeguard national security, strengthen biosecurity protections, and protect human health globally.
Second, supporting orderly international cooperation in China’s biomedical sector in accordance with laws and regulations.
The proposed controls include exemptions or facilitation measures for important forms of international cooperation.
For example, overseas licensing of relevant proprietary technology platforms and processes, as well as the provision of outsourced R&D and manufacturing services using those technologies, will be excluded from the scope of controls.
For proprietary technologies involving the discovery and validation of relevant therapeutic targets, exports conducted through overseas licensing arrangements will be eligible for facilitation measures such as automatic licensing.
Third, advancing the vision of a global community of health for all.
The proposed policy explicitly provides for automatic licensing and other facilitation measures for exports of restricted technologies required for public health emergencies, humanitarian assistance, or joint R&D using overseas resources.
China will continue to share the benefits of technological development with other countries and contribute to improving global health and well-being.
A: The proposal would add 12 photovoltaic technologies covering polysilicon manufacturing, silicon wafer manufacturing, and solar cell manufacturing.
The main considerations are as follows.
First, safeguarding national security and interests.
In recent years, photovoltaic technologies have increasingly been applied in sensitive areas such as aerospace, national defense, unmanned systems, and high-security infrastructure.
The potential national security risks associated with photovoltaic technology exports have therefore increased, making timely and proportionate regulation necessary.
Second, supporting the sharing of green technologies.
The scope of technologies proposed for control is limited. Technologies that do not meet the specified control thresholds may continue to be exported freely.
The announcement identifies a small number of key photovoltaic equipment categories. Controls apply only where such equipment incorporates technologies meeting the specified control thresholds.
Other photovoltaic equipment not explicitly listed in the announcement will, in principle, not be subject to its requirements.
Third, improving the technology export regulatory framework.
Exports of key photovoltaic technologies often take place alongside the export, installation, commissioning, modification, and upgrading of photovoltaic production lines.
Exporters engaged in such activities must apply for technology export licenses in accordance with applicable regulations and use the relevant documentation to complete foreign exchange, banking, tax, customs, and other procedures.
A: To better understand demand for technology exports, improve evidence-based regulatory management, and prevent potential export risks, the temporary registration requirement has been introduced with the necessary approval pursuant to Articles 18 and 20 of the Foreign Trade Law of the People’s Republic of China.
From the publication date of MOFCOM and MOST Announcement No. 48 of 2026 until the revised Catalogue formally takes effect, technology exporters must register exports of the technologies listed in the announcement through the Technology Trade Management Information System on MOFCOM’s Unified Business Platform before proceeding with the exports.
Exporters are responsible for determining whether their proposed technology exports fall within the scope of the temporary registration requirement.
Exports that have already commenced but have not yet been completed are also subject to registration.
The following circumstances require registration:
Transfers of technology ownership to overseas entities.
Licensing of domestically held technologies for use overseas.
Changes whereby the owner or actual controller of domestically held technologies becomes an overseas entity.
Other circumstances explicitly designated by MOFCOM as requiring registration.
It should be particularly noted that exports of key and core photovoltaic equipment falling within the scope of the Catalogue must also be registered in accordance with the relevant requirements.
Enterprises encountering difficulties or questions during the registration process may consult MOFCOM and MOST for assistance.
No posts

source