en.Wedoany.com Reported - Recently, China's Ministry of Industry and Information Technology (MIIT) issued the plan for the fourth batch of industry standard revision projects in 2026, officially approving 13 AI industry standards proposed by the MIIT's Artificial Intelligence Standardization Technical Committee. This batch of projects will advance standard development around AI technology, products, services, and governance needs, covering four sectors: security governance, product services, key foundational technologies, and basic commonalities, forming a clear task arrangement for subsequent technical specification compilation, industry opinion solicitation, and standard system construction.
This approval incorporates directions such as large models, intelligent agents, embodied intelligence, heterogeneous computing, data closed loops, industry adaptation, and technology ethics into the same batch of industry standard construction plans. The relevant content involves not only AI models and intelligent systems themselves but also extends to supporting links such as computing resources, data processing, industry applications, and security governance. The scope of standard development has expanded from single algorithms or products to multiple technical elements involved in the operation of AI systems.
In the direction of large models, standard construction will target representative technical forms in current AI products and services. As large models gradually enter different business scenarios, a more unified technical basis needs to be formed among model capabilities, product services, and industry applications. This batch of industry standard approvals will provide a project foundation for sorting out relevant technical requirements and subsequent specification compilation.
Intelligent agents and embodied intelligence are also listed as key directions. Intelligent agents involve the understanding, execution, and coordination of tasks by AI systems, while embodied intelligence further connects AI capabilities with physical devices, real-world environments, and specific operational scenarios. Both types of technology require relatively unified industry requirements in product forms, system capabilities, and application processes. This batch of standards includes these directions in the development scope, indicating that the AI standard system will continue to extend towards more complex system-level applications.
Heterogeneous computing is one of the key foundational technologies supporting the operation of AI systems. This approval advances it simultaneously with application directions such as large models and intelligent agents, reflecting that the construction of AI industry standards no longer focuses solely on model outputs and software functions but also begins to cover the organization, adaptation, and support capabilities among different computing resources, enabling the infrastructure layer and the model application layer to gradually connect within a unified standard framework.
Data closed loops and industry adaptation directly relate to whether AI systems can enter specific business environments. Data closed loops run through links such as data usage, model operation, and application feedback, while industry adaptation requires targeted adjustments to AI products and solutions based on business needs in different fields. This batch of standards includes both items in its key scope, helping to extend AI technical requirements from general capabilities to specific industry application processes.
Security governance and technology ethics constitute another important main line in this batch of standard projects. As the AI industry expands its application scope, it needs to simultaneously establish governance requirements that match technological development, ensuring that product R&D, service provision, and industry applications have clear regulatory foundations. This standard approval arranges security governance and technology ethics together with key technologies and product services, forming a standard development structure that advances technical construction and regulatory management in parallel.
From an overall layout perspective, the 13 industry standards are not isolated individual tasks but a combined standard construction effort centered around different links in the AI industry chain. Basic commonality standards are used to unify general requirements, key foundational technology standards target underlying support such as computing and data, product service standards connect model capabilities with practical applications, and security governance standards regulate the usage boundaries of related technologies and products. These four sectors together constitute an important part of the AI industry standard supply.
After the formal approval of the standards, subsequent work will enter the specific development phase. The MIIT's Artificial Intelligence Standardization Technical Committee will coordinate the advancement of the compilation of relevant industry standards and organize upstream and downstream enterprises in the AI industry, research institutes, universities, industry organizations, and relevant experts and scholars to participate, ensuring that the standard content can absorb opinions from multiple parties including technology R&D, product construction, industry applications, and research institutions.
The participation of entities covering different positions in the industry chain will also be an important feature of this batch of standard development. Enterprises can provide product R&D and application practices, universities and research institutes can participate in technical issue research, industry organizations can offer opinions based on industrial operation needs, and experts and scholars can contribute to standard content demonstration from a professional perspective, thus forming a multi-party participation mechanism for standard compilation.
As the 13 projects successively enter the development phase, China's AI industry standard system will further cover models, intelligent systems, computing, data, industry applications, and governance requirements. The focus of subsequent work will shift from project approval to standard content compilation, technical requirement sorting, and industry co-construction, ultimately forming a unified technical basis for reference in AI product R&D, service construction, and industry applications.
