Ministry of Industry and Information Technology
Cyberspace Administration of China
National Development and Reform Commission
Ministry of Education
Ministry of Commerce
State-owned Assets Supervision and Administration Commission of the State Council
State Administration for Market Regulation
National Data Administration
Notice on Issuing the Implementation Opinions on the “Artificial Intelligence + Manufacturing” Special Action
No. 279 [2025] of the Ministry of Industry and Information Technology
To the departments in charge of industry and information technology, the Cyberspace Administration offices of Party committees, development and reform authorities, education departments (education commissions), commerce authorities, state-owned assets supervision authorities, market regulation administrations (bureaus, departments, or commissions), and data management departments of all provinces, autonomous regions, municipalities directly under the Central Government, and cities with independent planning status, as well as the Xinjiang Production and Construction Corps, and all relevant entities:
Implementation Opinions on the “Artificial Intelligence + Manufacturing” Special Action are hereby issued to you. Please, in light of local conditions, conscientiously organize and ensure their effective implementation.
Implementation Opinions on the “Artificial Intelligence + Manufacturing” Special Action
Artificial intelligence (AI) is accelerating its in-depth integration with the real economy, profoundly changing the production models and economic forms of the manufacturing industry, and has become a key variable driving industrial upgrading and reshaping the global pattern. To implement the Opinions of the State Council on Deepening the Implementation of “Artificial Intelligence Plus” Initiative, accelerate the integrated application of AI technology in the manufacturing industry, foster new quality productive forces, and empower the new type of industrialization in an all-round, in-depth, and high-level manner, these implementation opinions are formulated.
I. General Requirements
Guided by Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era, we thoroughly implement the spirit of the 20th National Congress of the Communist Party of China and all plenary sessions of the 20th Central Committee, fully and accurately implement the new development concept, accelerate the construction of a new development pattern, coordinate development and security, and adhere to innovation-driven, scenario-led, market-oriented, safe and trustworthy, open and shared, and inclusive integration. We will focus on technological supply on one end to promote “intelligent industrialization”, and accelerate “industrial intelligence” on the other end by focusing on enabling applications, so as to comprehensively expand the industrial ecosystem, promote the in-depth integration of AI technological innovation and industrial innovation, and the “two-way empowerment” between AI technology and manufacturing applications. We will accelerate the intelligent, green, and integrated development of the manufacturing industry, and strongly support the construction of a manufacturing power, a cyber power, and a digital China.
By 2027, China will achieve safe and reliable supply of key core AI technologies, and its industrial scale and enabling level will rank among the top in the world. We will promote the in-depth application of 3-5 general large models in the manufacturing industry, form distinctive and full-coverage industry-specific large models, launch 1,000 high-level industrial agents, build 100 high-quality datasets in industrial fields, and promote 500 typical application scenarios. We will cultivate 2-3 ecologically leading enterprises with global influence and a number of specialized, sophisticated, and innovative small and medium-sized enterprises (SMEs), foster a group of enabling application service providers who “understand intelligence and are familiar with the industry”, and select 1,000 benchmark enterprises. We will build a world-leading open-source and open ecosystem, comprehensively improve safety governance capabilities, and contribute Chinese solutions to the development of AI.
II. Innovation Foundation: Consolidating the Enabling Base for Artificial Intelligence
(1) Strengthening Artificial Intelligence Computing Power Supply. Promote the coordinated development of hardware and software for intelligent chips, and support breakthroughs in key technologies such as high-end training chips, edge-side inference chips, AI servers, high-speed interconnection, and intelligent computing cloud operating systems. Orderly promote the layout of high-level intelligent computing facilities, accelerate the construction of a national integrated computing power network monitoring and dispatching platform, and promote the efficient utilization of computing power resources. Launch pilot projects for intelligent computing cloud services, promote the deployment of large model all-in-one machines, edge computing servers, and industrial cloud computing power, and enhance the supply capacity of intelligent computing resources.
(2) Developing High-Level Industry-Specific Models. Support innovations in model training and inference methods, and develop high-performance algorithm models adapted to the real-time, reliability, and security characteristics of the manufacturing industry. Cultivate key industry-specific large models, develop a “cloud-edge-end” model system, and continuously improve generalization capabilities. Build small models for segmented industrial scenarios and encourage collaborative innovation between large and small models. Promote the lightweight deployment of models and accelerate their application in industrial scenarios. Establish a public model service platform to provide high-level models and supporting tool services. Support the construction of a large model evaluation benchmark system, create authoritative rankings, regularly release evaluation results, and drive technological iteration and upgrading.
(3) Launching the “Model-Data Resonance” Initiative. Promote the establishment of a Chief Data Officer (CDO) system in enterprises, continue to promote the implementation of national standards for data management capability maturity, and consolidate the foundation of enterprise data governance. Sort out the list of data resources adapting to the needs of industry-specific models, issue guidelines for the construction of high-quality datasets in the manufacturing industry, make good use of carriers such as the Manufacturing Digital Transformation Promotion Center, and promote the transformation of basic data into high-quality industry-specific datasets to achieve “data guided by models”. Guide enterprises to strengthen the construction of data engineering capabilities, promote the in-depth integration of enterprise data development and model construction, explore the establishment of an integrated mechanism of “data collaboration, model training, application development, and security assurance”, and realize “models empowered by data”.
III. Intelligence Empowerment and Upgrading: Expanding and Promoting High-Value Application Scenarios
(4) Accelerating Enabling Applications in Key Industries. In-depth carry out the “In-depth Action” for AI empowering the new type of industrialization, and organize enabling service teams composed of high-level experts, enterprises, research institutions, etc. to go deep into industries, localities, and parks. Build an AI application docking platform to promote accurate matching of supply and demand. With reference to the “Guidelines for the Transformation of Key Manufacturing Industries Empowered by Artificial Intelligence” (see Annex 1), formulate industry-specific application panoramas and transformation roadmaps for “Artificial Intelligence + Manufacturing” by category, accelerate the empowerment of key manufacturing-related industries such as raw materials, equipment manufacturing, consumer goods, electronic information, and software and information technology services, and speed up the promotion and application of benchmark solutions and experiences.
(5) Accelerating the Transformation and Upgrading of the Entire Process. Systematically sort out application scenarios in key links, deepen the gradient cultivation of smart factories, promote the deep integration of large model technology into core production and manufacturing links, transform the entire process including R&D and design (including industrial design), pilot scale verification, production and manufacturing, marketing services, and operation management, and improve capabilities such as auxiliary design, simulation model construction, production scheduling, and predictive equipment maintenance.
- R&D and Design Link: Focus on promoting intelligent auxiliary design, auxiliary software code writing, drug R&D, etc., to create a new R&D and design model that is personalized, low-cost, and high-efficiency. Strengthen the construction, open-source, and sharing of industrial R&D datasets, explore the establishment of an AI prediction result evaluation system, improve engineering and technological innovation capabilities, and unblock the “barrier lake” of scientific discovery driven by AI.
- Pilot Scale Verification Link: Vigorously promote the intelligent transformation of pilot scale verification, accelerate the application of technologies such as virtual simulation and multi-modal fusion in the pilot scale verification link, and optimize process flows, improve pilot efficiency, and reduce test costs through comprehensive perception, real-time analysis, scientific decision-making, and precise execution.
- Production and Manufacturing Link: Deepen the application of AI technology in key industrial process control, process optimization, production scheduling, and other links, and promote the intelligence of production process analysis, decision-making, and execution. Popularize industrial quality inspection technologies such as machine vision and unmanned intelligent inspection, strengthen real-time monitoring of production lines and predictive maintenance, improve the accuracy of equipment fault identification, and realize early warning of safety production risks and event alarms.
- Marketing Service Link: Promote intelligent customer service, digital humans, and 3D product models, focus on breaking through personalized recommendations, customized after-sales services, and service-oriented extensions, develop functions such as consultation and training based on AI technology, improve the pre-sales, in-sales, and after-sales service experience, and enhance service value.
- Operation and Management Link: Give play to the reasoning and prediction capabilities of large models, accelerate the intelligent upgrading of links such as order processing, sales forecasting, and inventory early warning, and optimize supply chain management. Use the analysis and generation capabilities of large models to improve enterprises’ management capabilities in strategy, human resources, finance, risks, etc.
(6) Improving the Application Level of Key Enterprises. Carry out the evaluation of the intelligent maturity of manufacturing enterprises, implement the “Guidelines for the Application of Artificial Intelligence in Manufacturing Enterprises” (see Annex 2), and provide implementation paths and method guidelines for enterprises’ intelligent transformation and upgrading. Encourage leading enterprises, central state-owned enterprises (SOEs), etc. to take the lead in trials, provide large-scale application scenarios, develop and apply industrial agents, and take the lead in exploring new models of AI empowering the manufacturing industry. In-depth implement the special action for digital empowerment of SMEs, support SMEs in carrying out digital and intelligent transformation, and accelerate the replication and promotion of AI applications among SMEs.
(7) Promoting the Promotion and Application in Key Regions. Build and open a number of “Artificial Intelligence + Manufacturing” application scenarios, and create innovation highlands with industry characteristics. Relying on the advantages of resource agglomeration and talent concentration in National Independent Innovation Demonstration Zones, National High-Tech Industrial Development Zones, and National Economic and Technological Development Zones, accelerate the large-scale landing of new AI products, services, and business formats. Support advanced manufacturing clusters, digital industrial clusters, etc. in carrying out AI enabling applications, and promote the intelligent transformation and upgrading of regional manufacturing industries.
(8) Promoting the Intelligent Upgrading of Key Fields. Strengthen the coordination between AI and information and communication networks, promote the integrated empowerment of AI and industrial Internet platforms, develop datasets, large models, and agents for infrastructure such as the industrial Internet, and promote the in-depth application of AI technology in infrastructure planning, construction, operation, maintenance, and other links. Deepen the integrated application of AI technology in the field of green manufacturing, and develop and promote intelligent and green coordinated solutions for scenarios such as energy and carbon emission management and resource recycling. Build a number of application security solutions for industries, accelerate the landing application of security large models and agents, build a safe operation system, and improve the security level in the industrial field.
IV. Product Breakthroughs: Building New Intelligent Products and Business Formats
(9) Promoting the Iteration of Intelligent Equipment. Accelerate the integration and application of agents in various industrial equipment such as industrial mother machines and industrial robots, develop a new generation of AI numerical control systems, and improve capabilities such as independent decision-making, analysis, and execution. Accelerate the development of surgical robots, intelligent diagnosis systems, etc., and accelerate the innovation and clinical application promotion of intelligent medical equipment products. Promote the integration of AI technology into the R&D, manufacturing, and operation of major technical equipment such as large aircraft and ships, and develop intelligent low-altitude equipment such as unmanned aerial vehicles (UAVs). Conduct product testing and safety evaluation of intelligent connected vehicles equipped with autonomous driving functions, and steadily promote pilot projects for product access and road operation.
(10) Accelerating the Upgrade of Intelligent Terminals. Support technological breakthroughs in edge-side models and development application toolchains, and cultivate AI terminals such as smartphones, computers, tablets, and smart home appliances. Focus on key scenarios such as industrial inspection and telemedicine, and accelerate the industrialization and commercialization process of new terminals such as augmented reality/virtual reality (AR/VR) wearable devices and brain-computer interfaces. Promote the innovation of embodied intelligent products, build pilot scale verification bases and training grounds for humanoid robots, create benchmark production lines for humanoid robots, and take the lead in applying them in typical manufacturing scenarios.
(11) Creating a New Business Format of Agents. Carry out technological research on task planning and group collaboration of industrial agents, strengthen the integration of industrial mechanisms and agent decision-making models, and the interaction and adaptation between agents and industrial systems, and promote the cloud-based deployment of agents. Develop open and collaborative agent protocols and interfaces to improve the efficiency of interconnection and interoperability of agents. Support the construction and operation of agent application stores, select a number of typical cases of industrial agent applications, issue enterprise-level application practice guidelines, and accelerate the large-scale and commercialization process of agents. Build a classification and grading management system for agents, study the architecture of the agent Internet, explore mechanisms for agent registration and discovery, identity authentication, and access management, and guide the healthy development of new business formats. Accelerate the upgrading of traditional software products and services, promote the in-depth integration of AI and industrial software, and improve design and production efficiency.
V. Subject Cultivation: Building a Main Force for AI Development and Enabling Applications
(12) Gradient Cultivation of Enterprises. Support enterprises to increase R&D investment, actively undertake national major tasks, and gather resources to build ecologically leading enterprises with global influence. Develop AI enterprise incubators, implement SME entrepreneurship support programs, and gradually cultivate more specialized, sophisticated, and innovative “little giant” enterprises, high-tech enterprises, manufacturing single champion enterprises, unicorn enterprises, and gazelle enterprises in the AI field. Encourage local governments to provide enterprises with support such as “computing power vouchers” and “model vouchers”, strengthen public services for empowering SMEs, and reduce the cost of enterprise development and application.
(13) Building Innovation Carriers. Establish a National Manufacturing Innovation Center in the AI field to improve the supply capacity of key common technologies. Layout a number of key laboratories in the AI field to strengthen the exploration of cutting-edge technologies such as brain-inspired intelligence and world models. High-quality construction of national AI application pilot scale verification bases for key manufacturing industries, gather industrial innovation resources, and accelerate the formation of a number of replicable and promotable industry solutions.
(14) Developing Enabling Application Service Providers. Improve the digital and intelligent transformation service system for the manufacturing industry, build a number of AI enabling application accelerators, cultivate high-quality enabling application service providers, create enabling solutions combining standardization and customization, and provide services such as industry-specific model tuning, data governance, and security assurance. Encourage industrial enterprises, AI enterprises, and industrial Internet enterprises to gather resources such as tools, technologies, and platforms to build ecological partner service providers. Support telecommunications operators and digital intelligence technology companies of central SOEs to improve their service capabilities and undertake industry enabling application services. Guide relevant industry organizations to regularly release catalogs of high-quality service providers.
VI. Ecosystem Expansion: Strengthening Resource Allocation and Optimizing the Industrial Ecosystem
(15) Strengthening Standard Leadership. Give play to the role of technical standardization organizations such as the National Technical Committee for Artificial Intelligence of the Ministry of Industry and Information Technology (MIIT), the National Committee for Standardization of Data, the Subcommittee on Artificial Intelligence of the National Information Technology Standardization Technical Committee, the Artificial Intelligence Chip Working Group of the National Technical Committee for Integrated Circuits, and the Special Working Group on New Technology Security Standards of the National Technical Committee for Information Security, and strengthen the construction of standard technical organizations. Strengthen cross-industry and cross-field collaboration, and promote the development of basic standards such as safety, governance, and ethics, general standards such as software-hardware collaboration, enabling application standards, and measurement technical specifications by classification and grading. In-depth carry out the “AI Standards Promotion” activity to strengthen the promotion and application of standards. Encourage enterprises to participate in international standardization work.
(16) Promoting Open-Source and Openness. Build a high-level AI open-source community, deploy and implement a number of high-quality open-source projects such as models, datasets, and agents, and build a globally influential open AI ecosystem. Develop and promote open-source license agreements adapted to the characteristics of AI projects, and build a new open-source rule order for AI. Guide cloud service providers and enabling application service providers to actively connect with open-source communities, and promote the landing application of open-source projects in the industrial field. Organize activities such as developer conferences and “Campus-Source Tours” to spread open-source concepts, prosper open-source culture, and form a good atmosphere of co-construction and sharing.
(17) Strengthening Talent Introduction and Cultivation. Carry out demand forecasting for AI industry talents, release talent demand forecasting reports, and support universities and research institutes to layout in advance and adjust and optimize relevant disciplines and majors. Make good use of platforms such as the Beijing Zhongguancun College, Shanghai Chuangzhi College, Shenzhen Hetao College, National AI Industry-Education Integration Innovation Platform, National Outstanding Engineers College, and National Outstanding Engineers Practice Base, set up professional courses, cultivate compound talents who understand both AI and manufacturing applications, improve AI cognitive education and training, and enhance the AI literacy and skills of all employees. Strengthen the training of high-skilled talents in the AI field, cultivate scientific and technological leaders and innovation teams relying on national talent projects and programs, build a new model for cultivating leading talents in an unconventional way, and actively introduce high-end overseas talents.
VII. Safety Escort: Building a Safety Guarantee for Application Empowerment
(18) Improving Safety Guarantee Capabilities. Tackle key technologies such as deep synthesis authentication, industrial model algorithm security protection, training data protection, adversarial example detection, and intelligent terminal security evaluation, strengthen data security management, and enhance AI security protection capabilities. Build resources such as safety risk databases and corpora, and establish industrial safety large models. Enhance the transparency and interpretability of AI and reduce the risk of hallucinations through knowledge base optimization, training corpus error correction, and generation of synthetic content labels. Implement the Measures for the Administration of Ethical Management of Artificial Intelligence Technology, strengthen industry self-discipline, and improve enterprises’ ability to prevent AI ethical risks.
(19) Establishing a Safety Governance Mechanism. Study and formulate safety policies and standards such as classification and grading, evaluation and testing, and emergency disposal of AI in the field of industry and information technology, and support local competent departments to explore flexible governance mechanisms. Establish technical capabilities for AI safety risk monitoring and early warning, and strengthen risk monitoring, judgment, and prevention. Formulate guidelines for the reporting and sharing of AI safety risk information in the field of industry and information technology, coordinate the forces of all links in the industrial chain, and strengthen information sharing, risk notification, and collaborative disposal.
VIII. International Cooperation: Shaping New Advantages in International Cooperation and Competition
(20) Supporting Industrial Cooperation. Encourage enterprises to customize AI products and enabling application solutions according to the characteristics of different countries and regions. Launch the “Overseas Version” of the In-depth Action for AI Empowering the New Type of Industrialization, support industry organizations and professional institutions to provide supporting services for enterprises going global, guide enterprises to efficiently carry out various technical verifications and compliance certifications, and better serve the orderly overseas development of the industry. Guide foreign investment into the AI field, and encourage foreign-invested enterprises to carry out the development of generative AI technology and product production.
(21) Building International Cooperation Platforms. Actively participate in discussions on AI issues under cooperation mechanisms such as BRICS, the Shanghai Cooperation Organization (SCO), China-ASEAN, the Group of Twenty (G20), and the Asia-Pacific Economic Cooperation (APEC). Support the organization of high-end global influential competitions, exhibitions, and conferences such as the World Artificial Intelligence Conference and the Humanoid Robot Games in accordance with regulations, and actively promote China’s AI benchmark cases. Build the China-BRICS Artificial Intelligence Development and Cooperation Center with high quality, improve the level of practical cooperation, and promote the coordinated development of the global industry.
IX. Guarantee Measures: Strengthening Comprehensive Policy Support and Guarantee
Establish a work promotion mechanism of inter-departmental cooperation, central-local linkage, and industrial collaboration, encourage local governments to formulate policies and measures according to local conditions, guide enterprises to develop in a differentiated manner, and prevent “involutionary” competition in the industry. Coordinate existing funding channels to support technical R&D and enabling application tasks related to “Artificial Intelligence + Manufacturing”. Give play to the role of the National Artificial Intelligence Industry Investment Fund, enrich the reserve of high-quality projects, and attract more social capital to increase investment in an orderly manner. Launch large-scale application demonstration actions for new technologies, new products, and new scenarios, make good use of policies for the application of first-of-its-kind (FOIK) equipment, first-batch products, and first-version software, promote the promotion, application, iteration, and upgrading of new technologies and new products, and release the potential of domestic market demand. Carry out the calculation of the scale of the AI industry, establish an application monitoring and evaluation indicator system, improve the AI industry monitoring and analysis platform, and dynamically monitor the development trend of the global industry.
Annexes:
1. Guidelines for the Transformation of Key Manufacturing Industries Empowered by Artificial Intelligence
2. Guidelines for the Application of Artificial Intelligence in Manufacturing Enterprises
Annex 1:
Guidelines for the Transformation of Key Manufacturing Industries Empowered by Artificial Intelligence
The manufacturing industry is the mainstay of the national economy, the foundation of building a country, the instrument of rejuvenating the country, and the cornerstone of making the country strong. The in-depth integration of AI and the manufacturing industry is an important path to develop new quality productive forces and build a modern industrial system. To give play to the advantages of the manufacturing industry in large scale, complete categories, and rich scenarios, and in combination with the characteristics, technical maturity, digitalization level, and other basic conditions of various industries, we will promote the application of AI in the manufacturing industry by classification and implementation, and accelerate the intelligent, green, and integrated development of the manufacturing industry. These guidelines are formulated accordingly.
I. Raw Materials Industry
(I) Improving the Whole-Process Intelligent Level of the Iron and Steel Industry. Build public products such as datasets and knowledge bases for the iron and steel industry, create an engineering application platform for AI, and provide intelligent solutions. Develop a series of dynamic models covering the entire iron and steel production process, and based on iron and steel mechanism knowledge and production practice experience, develop large models and agents for the iron and steel industry in areas such as vision, prediction, and decision-making, so as to realize real-time perception of the operating conditions of key equipment, adaptive optimization of process parameters, product performance prediction, quality defect traceability, global optimization of scheduling tasks, and real-time intelligent adjustment. Promote AI to empower the entire process of the iron and steel industry, and improve production efficiency, product quality, resource efficiency, safety, and service levels.
(II) Promoting Quality Improvement and Efficiency Enhancement in the Petrochemical and Chemical Industry. Comprehensively use large models and digital twin technologies to break through the R&D paradigm of oil and gas exploration and development and chemical new materials. Deeply integrate process mechanisms, expert experience, production and operation data, etc. of oil and gas production operations, pipeline storage and transportation, and chemical processes, build large models for the petrochemical and chemical industry, promote the integrated application of large and small models, and realize safety monitoring and early warning of oilfield operation areas and chemical production, predictive equipment maintenance, adaptive optimization of process flows, product quality prediction, etc. Build data infrastructure such as high-quality industry datasets and data resource nodes to support the training and development of industry-specific large models and agents, and improve the level of AI application in complex scenarios.
(III) Accelerating the In-Depth Integration of AI and New Materials R&D. Establish a new materials big data center, build a high-precision, long-sequence, and multi-modal materials industry dataset, and improve the standardization level of industry data formats. Develop a cross-scale computing framework for alloys, ceramics, polymers, energy materials, etc., build industry-specific large models for new materials molecular design, synthesis and preparation, process optimization, etc., and improve the reverse design capability of material “composition-structure-performance”. Establish a large model prediction result evaluation system to enhance the accuracy of model prediction. Improve the human-machine collaboration capability in materials science research, and enhance the high-throughput automated experiment and preparation capability of new materials.
(IV) Promoting AI to Empower the Non-Ferrous Metals Industry. Develop technologies and tools for automated data governance and annotation, build high-quality industry datasets such as mine and equipment operation, mineral processing process optimization, and smelting process control, and build a data foundation support system. Establish non-ferrous metals industry-specific large models, scenario models, and agents integrating “physical mechanisms-process data-environmental variables”, promote the collaborative application of large and small models, meet the use requirements such as reliability and dynamic adaptability, and realize the innovation of R&D models for new materials and new processes, precise control of mining, mineral processing, and smelting processes, real-time optimization of key parameters, and accurate classification and identification of recyclable resources.
(V) Promoting the Innovative Application of AI in the Building Materials Industry. Prioritize the deployment of scenario models for the typical unit operation needs of industries such as cement and flat glass, train and build large models for the building materials industry, and promote their in-depth application in scenarios such as mining, raw material ratio optimization, kiln calcination control, and cement clinker strength prediction, so as to improve the level of intelligent optimization and control of the production process. Promote the development of an intelligent algorithm system of “data-driven + mechanism model”, build datasets for advanced inorganic non-metallic materials such as advanced ceramics and artificial crystals, and promote the development of new products and the optimization of production processes.
II. Equipment Manufacturing Industry
(I) Promoting the Flexible and Intelligent Leap of Industrial Mother Machines. Deeply integrate AI technology into numerical control systems to empower the entire process of “real-time perception-autonomous learning-intelligent decision-making-closed-loop execution”, and improve the adaptive operation and execution capabilities of industrial mother machines. Build an intelligent diagnosis system based on large models to accurately perceive and judge equipment status, and realize remote monitoring and predictive maintenance. Relying on modular production units and intelligent decision-making services, through low-code configurable task scheduling and autonomous resource scheduling, realize the autonomous response of the manufacturing system to order changes, real-time reconstruction of production lines, and agile production.
(II) Accelerating the Whole-Chain Intelligent Upgrade of the Automobile Industry. Build automotive large models to automatically generate schemes such as body shape and interior layout, dynamically optimize parameters such as structural strength and drag coefficient through real-time simulation, and promote a new paradigm of intelligent R&D. Accelerate the application of AI technology in links such as hardware configuration and parameter tuning, develop modular process islands, and build flexible and reconfigurable production lines. Establish AI-driven whole-process quality control and predictive maintenance, and promote online testing of vehicle performance and full-life-cycle quality traceability.
(III) Promoting the Whole-Life-Cycle Intelligence of Power Equipment. Based on AI technology, intelligently optimize the structural parameters of core components such as generators, and promote the digital twin design and test simulation of large power generation equipment. Use AI algorithms to strengthen the manufacturability analysis of power equipment and intelligently evaluate the processing difficulty of components and assembly compatibility. Build an AI-driven health assessment and life prediction platform, carry out condition-based maintenance, and improve the level of intelligent monitoring and scheduling optimization of power generation and transmission equipment.
(IV) Promoting the Application of AI Technology in the Shipbuilding Industry. Build large models for the shipbuilding industry, explore new R&D and design models, and promote the intelligent upgrading of key processes such as “cutting-welding-spraying-logistics” to meet the needs of unmanned and intelligent production of large ships and marine equipment. Promote the construction of AI application scenarios in fields such as marine equipment manufacturing and smart ports. Based on AI technologies such as data governance and machine learning, establish an operational performance model of ship equipment systems to realize functions such as ship navigation energy efficiency optimization and equipment fault diagnosis.
(V) Building an Intelligent Manufacturing System for Aerospace. Develop a simulation platform based on AI algorithms, combine aerodynamic data and computational fluid dynamics simulation models to automatically iterate schemes such as fuselage lines and airfoil profiles, realize extreme working condition verification, and shorten the test cycle. Build an industrial decision-making system and carry out the application of agents in links such as design, manufacturing, operation and maintenance, and management. Construct AI solutions such as intelligent processing and assembly of large and complex material components, additive manufacturing and intelligent inspection of special materials, and intelligent integration and testing of spacecraft final assembly, so as to comprehensively improve the intelligent level of the industry.
III. Consumer Goods Industry
(I) Improving the Capability of Personalized Design and Efficient Production in the Textile and Apparel Field. Build an intelligent product planning platform for the apparel industry, deeply mine massive consumer data, and use data analysis and decision-making large models to realize the rapid identification of apparel product hotspots and the design of response schemes. By integrating physical engines and 3D generative large models, build a personalized design and virtual fitting system to improve the consumer shopping experience. Promote the deployment of adaptive production systems to realize micron-level yarn tension monitoring and defect self-repair, and improve product yield. Develop intelligent sorting technology and equipment for waste textiles based on multi-spectral intelligent identification to improve the utilization rate of renewable resources.
(II) Strengthening the Intelligent Operation and Intelligent Product Supply Capability in the Home Furnishing Field. Establish a data-driven product design agent to optimize product structure and functions, improve intelligent control capabilities, and accelerate the pace of new product launches. Integrate industrial scheduling large models with industrial Internet technology, connect multi-source data such as production equipment, orders, and materials, realize collaborative scheduling of multiple production lines and warehouse scheduling, and enhance manufacturing flexibility and response speed. Develop smart home products with functions such as human-machine interaction, intelligent perception, and intelligent interconnection, build diverse scenarios, establish an active service-oriented home appliance reminder system, provide energy-saving schemes and predictive maintenance, and improve equipment operation reliability and user satisfaction.
(III) Building a Safe, Efficient, and Intelligent Management System in the Food Processing Field. Encourage the application of AI technology to enrich the supply of AI large model products in the food industry. Organize food enterprises and professional service providers to provide intelligent solutions such as intelligent monitoring and traceability of food production, “5G + Industrial Internet” in food parks, and intelligent management of raw material production and supply. Accelerate the R&D and deployment of multi-modal safety production monitoring large models to improve the real-time identification capability of irregular operations and dangerous behaviors in food production sites. Improve supply chain risk prediction and emergency response capabilities, real-time perceive supply chain disruption risks, and ensure the stability of food supply.
(IV) Improving the Level of Intelligent Pharmaceutical R&D and Supply Chain Management. Build an AI-driven new drug discovery and virtual screening platform, and accelerate target identification and candidate drug discovery through multi-modal drug efficacy prediction large models, so as to reduce the drug R&D cycle and cost. Integrate quantum chemical simulation with AI technology to accurately design drug molecular structures and improve drug efficacy and safety. Accelerate the application of AI in links such as drug synthesis route planning and raw material combination optimization, and build an automated, high-throughput, and low-cost intelligent drug synthesis system. Establish an intelligent pharmaceutical supply chain management platform to real-time track changes in drug demand, dynamically optimize inventory and distribution routes, and avoid drug shortages and waste.
(V) Promoting the Whole-Chain Innovative Development of the Biomanufacturing Field. Use AI technology to mine and generate high-performance biological components, efficient synthetic metabolic pathways, and high-activity enzyme protein structures, and enrich basic databases. Build an intelligent strain construction platform to accurately simulate the operation mechanism of cell factories and create high-conversion industrial strains. Establish a prediction model between process parameters and product yield to shorten the process development cycle and improve the success rate of pilot scale verification. With the help of AI and other technologies, optimize and iterate core parameters such as temperature, pH value, and oxygen content in the biological reaction process, realize intelligent control of the reaction process, and accelerate the industrialization process.
(VI) Promoting the Renewal and Upgrade of Historical and Classic Industries. Accelerate the construction of a brain for historical and classic industries, build an industrial data base integrating core technologies such as silk patterns, porcelain glaze formulas, and tea roasting processes, and realize the accurate connection between market demand perception and product innovation. Relying on technologies such as AI and the industrial Internet, promote customized and collaborative design innovation, and drive the transformation of cultural IP into fashion consumer goods. Use technologies such as machine vision to build a whole-process quality control system, reproduce the production scenes and process flows of classic industries through 3D modeling and digital twin technology, and create an immersive cultural space integrating technology display, interactive experience, and customized production to improve the consumer shopping experience.
IV. Electronic Information Industry
(I) Improving the Intelligent Design Level of Electronic Components. Realize full virtual simulation and debugging of electronic components through generative AI and digital twin technology, and build a cross-domain collaborative R&D platform. By integrating advanced computing engines and multi-modal large models, break down data silos between electronic design automation (EDA) and product lifecycle management (PLM) systems, and support the rapid iteration and verification of complex chip architectures and new display devices. Focus on breaking through high-precision simulation and prediction technologies for electronic components to shorten the R&D cycle and reduce physical trial-and-error costs.
(II) Promoting Flexible Intelligent Manufacturing in Industries Such as Consumer Electronics and New Displays. Realize the dynamic reconstruction of production lines through industrial large models and edge intelligence technologies, and build an adaptive flexible production system for the electronic information industry. Deploy AI-driven process parameter optimization models, combine machine vision with multi-scale physical property characterization, and realize millisecond-level tuning of key processes such as electronic component mounting, assembly, and testing. Develop modular and intelligent electronic information manufacturing equipment and low-latency networks to support multi-variety and small-batch production in the consumer electronics and new display industries, significantly reduce line change time, and improve overall equipment efficiency (OEE).
(III) Strengthening the Quality Control Capability of Electronic Information Components and Products. Accelerate the construction of a knowledge graph for the electronic information industry, realize intelligent analysis of quality root causes, and build a whole-process quality control platform. Develop online quality inspection systems covering links such as printed circuit board (PCB) design and chip packaging, integrate machine vision, non-destructive testing, and multi-spectral identification technologies, and improve the efficiency and accuracy of electronic component inspection. Establish a knowledge base and prediction model for quality defects of electronic information products, effectively reduce the defective product rate, improve the response speed of quality traceability, and promote the transformation from post-event remediation to proactive prevention.
(IV) Innovating Intelligent Solutions for Green and Low-Carbon Development in the Electronic Information Industry. Integrate AI and blockchain technology to realize the full-life-cycle accurate accounting and trusted data sharing of carbon footprints of electronic information products. Develop carbon management large models for the photovoltaic and lithium battery industries, integrate industrial Internet identification resolution with energy consumption prediction algorithms, and dynamically optimize equipment parameters and energy scheduling. Deploy intelligent power prediction and station operation systems to promote a significant reduction in energy consumption per unit output value of the energy electronics industry, improve the credibility of carbon emission data, and support the high-end extension of the global value chain.
V. Software and Information Technology Service Industry
(I) Building an Intelligent Toolchain Product System for the Whole Software Lifecycle. Focus on multi-modal large models, behavior analysis, time-series prediction, etc., and build an intelligent development toolchain product covering software requirement design, development, testing, and operation and maintenance. Create AI-driven development and operation (DevOps) products to realize intelligent scheduling and risk early warning. Cultivate vertical low-code platforms and agent development platforms, and realize the rapid encapsulation of industry knowledge, automated task design and execution through modular AI components, so as to promote the transformation of software development from “human-led” to “intelligent collaboration”.
(II) Accelerating the Intelligent Upgrade of Traditional Software and Services. Promote the integration of AI technology with basic software, industrial software, and manufacturing industry-specific application software to realize the intelligent upgrade and value reconstruction of traditional software. Improve the dynamic perception, self-optimization, and self-evolution capabilities of software, and realize the dynamic reorganization and performance optimization of software functional modules. Integrate technologies such as predictive analysis and business process mining to empower software with intelligent decision-making capabilities. Based on the domestic agent interconnection protocol, develop high-performance intelligent communication middleware to realize efficient collaboration between software and large models and unified analysis of multi-source data.
(III) Cultivating and Building Vertical Domain Agents. Develop and deploy agents for software programming, software requirement and audit, and software testing; build industrial agents for surface design, automatic modeling, automatic programming, etc.; develop specialized agents for intelligent scheduling and planning, dynamic report generation, interface automation design, and intelligent data monitoring and governance. R&D agents for industries such as medical care, education, finance, and law.
(IV) Building High-Quality Datasets for the Software Industry. Breakthrough technologies such as automated cleaning of multimodal data and intelligent semantic annotation, and build standardized software R&D datasets. Use synthetic data and adversarial testing technologies to simulate complex boundary scenarios such as high concurrency and network anomalies, and build real-scenario test datasets. Based on fine-grained entity relationship extraction and heterogeneous multi-source knowledge alignment technologies, build semantic domain knowledge assets. Establish an open-source code compliance cleaning pipeline to effectively filter license conflicts and vulnerability risks, and comprehensively consolidate the data base for the integrated innovation of “AI + software”.
Annex 2:
Guidelines for the Application of Artificial Intelligence in Manufacturing Enterprises
The in-depth integration of AI with all elements, processes, and chains of the manufacturing industry is an important way to break through the bottlenecks of industrial upgrading and shape international competitive advantages. To accelerate the in-depth integration of AI and the manufacturing industry, better combine digital technology with manufacturing advantages, improve the scientific and standardized level of AI application in manufacturing enterprises, and fully empower the new type of industrialization, these guidelines are formulated.
These guidelines apply to enterprises that use AI for R&D and design, production and manufacturing, operation and management, and extended services.
I. Conducting Intelligent Evaluation and Planning
(I) Conducting Diagnosis and Evaluation of Intelligent Level. Comprehensively use reference standards such as data management capability maturity, smart manufacturing capability maturity, digital transformation maturity, and the Integration of Informatization and Industrialization Management System, as well as the general evaluation indicator system for manufacturing digital transformation, to find out the digitalization, networking, and intelligentization levels of enterprises, and identify the bottlenecks of transformation and upgrading. Combined with economic analysis and risk assessment, scientifically determine AI application needs.
(II) Formulating AI Application Plans. With reference to typical application cases of AI empowering the new type of industrialization, determine the core scenarios of AI application and the priority of technology introduction, and reasonably set application goals. Prioritize the intelligent upgrading of scenarios such as operation and management, R&D and design, and gradually layout the transformation and upgrading of links such as pilot scale verification and production and manufacturing. Give play to the supporting role of the industrial Internet digital base, strengthen the overall connection with enterprise digital transformation work, and ensure that AI applications accurately support the development of core businesses.
II. Improving Intelligent Basic Capabilities
(III) Upgrading Hardware Basic Capabilities. Carry out digital transformation and upgrading of industrial “dumb equipment” and “dumb positions”, and build a hardware support system combining a unified technical base and scenario-based application suites. Comprehensively improve the information perception, transmission, decision-making, and control capabilities of various scenarios by installing sensing equipment and intelligent instruments, deploying edge computing equipment, promoting the upgrading of industrial private networks, and applying general tool products for digital transformation. Through the optimization and upgrading of computing, storage, and networks, accelerate the transformation of existing data centers into intelligent computing centers.
(IV) Improving the Intelligent Level of Software. Accelerate the intelligent transformation and upgrading of core software such as industrial real-time operating systems, control and optimization software such as manufacturing execution systems (MES) and online real-time optimization software, and control and execution units such as distributed control systems (DCS) and data acquisition and monitoring systems (SCADA), so as to improve intelligent supporting capabilities. Optimize the core of basic software, embed intelligent scheduling algorithms, improve resource allocation efficiency, and enhance system response speed. Deploy industrial software integrating digital twin, large models, and other digital intelligence technologies for industrial design, production control, operation and management, and service support, and strengthen the native intelligent foundation of industrial software.
III. Building High-Quality Datasets
(V) Building a Data Resource Platform. Establish an enterprise-specific knowledge database to form a data resource pool covering the entire business scenarios such as R&D and design, production and manufacturing, supply chain management, and operation and decision-making management. Build an industrial knowledge base including a mechanism library (storing underlying principle knowledge such as industrial mechanism models, technical documents, and design drawings), a simulation library (storing multi-disciplinary simulation models), and an experience library (storing practical knowledge such as fault cases, best practices, and operation skills) to effectively support the needs of enterprise AI datasets. Establish an integrated enterprise data management platform to support the aggregation, processing, annotation, and quality evaluation of multi-source heterogeneous data, improve the enterprise’s data processing and utilization capabilities, and enhance dataset quality.
(VI) Applying Dataset Processing Toolchains. Strengthen the use of data processing tools, gradually covering key links such as data aggregation, collection, cleaning, enhancement, annotation, synthesis, storage, transmission, analysis, and application, to continuously provide high-quality, efficient, and safe dataset support for enterprise AI applications. Focus on strengthening the use of tools in directions such as intelligent annotation, expert collaborative annotation, synthesis of mechanism and simulation data, dataset quality evaluation, and security monitoring.
(VII) Establishing a Data Management System. Encourage enterprises to explore the Chief Data Officer (CDO) system, establish a data management system covering planning, implementation, evaluation, and improvement, strengthen data standardization construction, and promote data integration among various systems. Establish a classification, stratification, and grading management mechanism for enterprise datasets, and comprehensively consider factors such as data type, data system, application scenario, and security to ensure the safe application and effective circulation of enterprise datasets. Clarify the key steps and quality points of links such as data collection, preprocessing, data annotation, enhancement and synthesis, and dataset productization, formulate dataset quality evaluation standards, and guide the improvement and efficient application of dataset quality.
(VIII) Building Diversified Datasets. Focus on links such as R&D and design, production and manufacturing, and operation and management in the industrial field, and build multi-modal high-quality industrial datasets covering enterprise scenarios such as process design optimization, process control, fault diagnosis, and intelligent operation. Encourage manufacturing enterprises to cooperate with third parties to build datasets such as synthetic datasets, deep thinking chain datasets in the industrial field, and cross-disciplinary and cross-domain knowledge graphs, create high-quality industry-specific datasets, explore dataset productization, and support industrial AI applications in complex scenarios.
IV. Reasonably Planning and Layout Computing Power Resources
(IX) Scientifically Planning Computing Power Scale. In accordance with the national overall deployment and combined with the actual development of enterprises, formulate a phased and gradually increasing computing power deployment scale, and encourage the priority selection of computing power services that can achieve real-time response and scalability.
(X) Reasonably Allocating Computing Power Resources. Encourage the priority use of cloud computing services to quickly build intelligent basic service capabilities and reduce technical investment costs. Enterprises with a good digital foundation and high requirements for data security can build intelligent computing resources based on their own computing power infrastructure, deploy AI applications, and realize intensive resource utilization.
(XI) Strengthening the Collaborative Scheduling of Computing Power Resources. Encourage enterprises to realize cloud-edge-end computing power collaboration based on business characteristics, integrate multi-source heterogeneous chip resources, realize tasks such as model training, fine-tuning, quantization, and distillation on the cloud side, and realize lightweight model deployment on the edge side to meet the low-latency requirements of industry. Dig deep into computing power usage needs and application scenarios, and deepen the connection between computing power supply and demand and the efficient scheduling and operation of computing power resources.
V. Conducting Model Selection and Tuning
(XII) Scientifically Determining Application Scenarios. Focus on solving key technical or process problems of enterprises in the entire manufacturing process, select high-value scenarios that can significantly drive productivity, carry out AI technology R&D and application landing, and focus on deploying AI applications in the following five types of scenarios: R&D and design focusing on intelligent auxiliary design and rapid generation of creative drawings; pilot scale verification focusing on intelligent construction of simulation models and intelligent generation of test data; production and manufacturing focusing on intelligent production scheduling and industrial vision intelligent inspection; marketing services focusing on personalized recommendations and customized after-sales services; and operation and maintenance management focusing on predictive equipment maintenance, energy efficiency optimization analysis, and auxiliary business decision support.
(XIII) Quantifying Key Scenario Indicators. Combine scenario characteristics and business goals to set quantifiable indicators for model selection, so as to evaluate scenario application effects and provide a basis for model selection and tuning. For R&D and design scenarios, focus on measuring indicators such as the number of design iterations per unit time, the number of design scheme generations, and the scheme adoption rate; for pilot scale verification scenarios, focus on assessing indicators such as simulation modeling efficiency and test indicator coverage; for production and manufacturing scenarios, focus on monitoring indicators such as comprehensive optimization efficiency, production qualification rate, and missed detection rate/false positive rate; for marketing service scenarios, focus on checking indicators such as marketing conversion rate and response time; for operation and maintenance management scenarios, focus on indicators such as fault prediction accuracy and maintenance cost reduction.
(XIV) Selecting Models Based on Business Needs. Based on business scenario requirements and combined with the construction of computing power infrastructure, carry out model evaluation and selection. Comprehensively consider the compatibility, reliability, and ease of use between models, development frameworks, compilers, and inference deployment toolchains, and give priority to mature solutions verified by industry practice. Encourage the R&D of agent products for segmented business scenarios in the manufacturing industry and the construction of intelligent solutions. Take safety as an important consideration for model selection, comprehensively consider factors such as model source, vulnerabilities and defects, and security protection mechanisms, and give priority to model bases with high safety and credibility. Encourage enterprises to build the collaborative capability of models across the entire chain of production, supply, and marketing to improve the linkage efficiency of various links.
(XV) Adopting Prompt Engineering and Retrieval-Augmented Tuning. Build a prompt library covering common industrial problems and edge cases, and establish multi-dimensional indicators such as grammatical correctness, semantic completeness, and user satisfaction. For high-frequency knowledge update scenarios such as market analysis and new technology application, connect to industry databases and information platforms, implement an authority evaluation and content monitoring mechanism for data sources, and ensure information authenticity.
(XVI) Using Model Fine-Tuning to Adapt to Typical Scenarios. For quality inspection and defect identification scenarios, focus on carrying out small-sample annotated defect data fine-tuning based on pre-trained models to enhance the model’s ability to extract complex and tiny features; for production scheduling scenarios, focus on full-parameter fine-tuning of time-series prediction models based on historical production line data to dynamically allocate resources and improve core task efficiency; for equipment fault diagnosis scenarios, focus on real-time monitoring and prediction using multi-modal data such as time-series data and audio data to optimize fault prediction models.
(XVII) Conducting Hybrid Tuning Based on Actual Conditions. Encourage enterprises to give priority to prompt engineering and retrieval-augmented technology according to actual conditions, and gradually try parameter-efficient fine-tuning and full-parameter fine-tuning to improve model capabilities. Build a multi-modal model candidate library based on actual conditions, and comprehensively adopt methods such as parameter fine-tuning, architecture search, and large-small model collaboration to determine the optimal solution.
VI. Model Deployment and Integration
(XVIII) Verifying Model Performance. Conduct trial operation verification in the actual production environment to ensure that the model can operate effectively in real scenarios. Comprehensive consideration of resource allocation, data security, real-time performance, stability, response capability, and system scalability requirements of various models, use technologies such as microservice architecture, API interfaces, and middleware to centrally deploy models or implement cloud-edge-end collaborative deployment based on business characteristics.
(XIX) Improving Model Usability. According to business needs, develop specific model application interfaces, low-code components, etc., and realize flexible configuration of data access and display of model analysis results based on business requirements.
VII. Continuously Improving Application Effects
(XX) Evaluating Application Capability Level. Set up a professional team to conduct special evaluations and regular analysis and improvement. Evaluate the problems of AI in enterprise applications from aspects such as model accuracy, computing power utilization rate, inference latency, investment cost, and safety and stability.
(XXI) Promoting Iterative Optimization and Upgrading. Regularly analyze the impact of AI application on the improvement of enterprise operation and decision-making levels, business processing efficiency, product production quality, and operating efficiency. Combine enterprise development strategies and AI technology trends to formulate application goals and implementation plans for the next stage. Strengthen intensive management and control, promote the in-depth integration of intelligence and greenization, and realize sustainable development.
(XXII) Deepening Technological Integration and Innovation. Cooperate with universities and research institutions to tackle key technologies such as real-time performance, edge-side deployment, and reliability of models in industrial applications. Based on application effects, promote secondary innovation and deeply embed industry-specific large models into the entire process of R&D and design, pilot scale verification, production, and operation. Strengthen parameter optimization and knowledge reasoning capabilities, incubate industrial intelligent software and hardware tools and products such as intelligent software development and intelligent operation and maintenance, and build new quality productive forces driven by AI.
(XXIII) Encouraging the Output of Excellent Solutions. Leading industry enterprises with technological advantages should output overall technical solutions to the upstream and downstream of the industrial chain by opening platform interfaces, open-sourcing general models and toolchains, sharing high-performance algorithm models, and formulating standards and specifications, so as to promote collaborative innovation in the industrial chain.
VIII. Ensuring AI Application Security Protection
(XXIV) Strengthening Data Security Protection. Implement laws and policies such as the Data Security Law and the Measures for the Administration of Data Security in the Field of Industry and Information Technology (Trial), and organize work such as data classification and grading, full-life-cycle security protection, risk monitoring and early warning, and risk assessment according to the characteristics of industry-specific data to provide data security guarantee for AI applications in various industries. For links such as data annotation, aggregation, training, and synthesis, strengthen data verification, detection and evaluation, identity authentication, and permission management to improve the level of data security risk prevention.
(XXV) Preventing Application Security Risks. For typical AI application scenarios such as R&D and design, pilot scale verification, production and manufacturing, marketing services, and operation and management, encourage enterprises to regularly carry out safety testing and evaluation of industrial large model hallucinations, accuracy, robustness, etc. Establish dual-end filtering and security monitoring capabilities for AI application input and output, and strengthen the prevention of risks such as malicious instruction input and abnormal reasoning output. Strengthen the security management of the AI application supply chain, including the security capabilities of upstream and downstream suppliers into the key points of partner management, and build a sound supply chain security governance capability.
(XXVI) Improving the Level of Network Security Protection. Promote the integration of network security into all links of AI planning, deployment, and application in manufacturing enterprises, implement the Cybersecurity Law and the Measures for the Classification and Grading Management of Industrial Internet Security, carry out work such as independent grading, information registration, classified protection, compliance evaluation, and security rectification, improve the enterprise network security management and protection system, strengthen the network security capabilities of industrial control systems, and improve the level of risk prevention in the process of AI application.
IX. Strengthening Organizational Guarantee
(XXVII) Implementing Enterprise Main Responsibility. Systematically formulate management systems for enterprise digital and intelligent transformation and upgrading, strengthen enterprise resource guarantee, and efficiently and steadily promote the in-depth development of AI applications.
(XXVIII) Strengthening the Training of Compound Talents. Strengthen industry-university-research-application collaboration, encourage universities and enterprises to support the training of top-notch innovative AI talents relying on platforms such as the National AI Industry-Education Integration Innovation Platform and demonstration characteristic colleges, improve the mechanism for introducing, evaluating, and incentivizing AI talents in enterprises, create a good talent development environment, and cultivate compound talents with both industry cognition and technical practical capabilities.
(XXIX) Actively Participating in Ecosystem Co-Construction. Timely summarize successful experiences, actively share AI solutions, build industry application benchmarks, and promote the improvement of the intelligent level of the manufacturing industry.
