China Releases First National Standard System for AI Agent Interconnection

The 2026 Global Digital Economy Conference, held from July 2 to 5 in Beijing, served as the venue for a major policy announcement in China's AI sector. Han Xia, Executive Deputy Director of MIIT's Information and Communication Technology Committee, stated at the Agent Internet Forum that 2026 marks a strategic window for large-scale deployment of the agent industry, with the agent internet gradually moving from concept to practice.

The newly released "Artificial Intelligence — Agent Interconnection" national standard series (GB/Z 185—2026) is China's first standard system dedicated to AI agent interoperability. According to MIIT, the implementation of this standard series will drive agents from single-point applications and closed systems toward cross-platform, cross-system, and cross-scenario collaboration, providing foundational support for large-scale industrial deployment.

The standard series builds a complete technical framework addressing key aspects of agent interconnection: trusted identity, capability visibility, discovery and matching, interaction and collaboration, and tool invocation. For trusted identity, the standards employ mechanisms including identity codes, identity accounts, identity credentials, and identity authentication to ensure security in cross-system agent collaboration.

Han Xia noted that the global agent internet is currently at a critical juncture where both technical and governance paradigms are being established. Accelerating the evolution of agents from individual intelligence to collective intelligence and from closed applications to open collaboration is crucial for shaping China's new AI development landscape. She also acknowledged persistent challenges, including unclear technical architecture, the need for network infrastructure upgrades, and the lack of unified identification and protocol standards.

Wang Zhiqin, Vice President of the China Academy of Information and Communications Technology (CAICT), emphasized that AI is transitioning from breakthroughs in individual capabilities to systematic applications. As various types of agents are deployed at scale, the demand for cross-platform, cross-entity, and cross-application interoperability continues to grow. She described the agent internet as a systemic project with broad coverage, a long technical chain, and pronounced cross-industry integration characteristics, still in an early exploration phase.

A long-standing pain point in the industry has been the lack of unified standards across different vendors, platforms, and architectures for agent identity recognition, capability description, and discovery matching, resulting in high adaptation costs and low collaboration efficiency. While international open protocols such as MCP and A2A have emerged, they have yet to form a unified consensus and cannot fully meet industrial requirements. China's national standard system could serve as an important reference for the evolution of global agent interconnection standards.

Fan Kefeng, Vice President of the China Electronics Standardization Institute, described the standards as establishing fundamental principles and capability thresholds for AI technology development and industrial application. The framework transforms isolated systems into interconnected networks, proprietary interfaces into open protocols, and disorderly development into orderly progression. At a critical stage where agent capabilities are rapidly improving and deeply engaging with physical scenarios, defining safe and compliant interaction boundaries in advance helps achieve a dynamic balance between innovation and governance.

The release is also part of a broader systematic push for AI standardization in China. According to the State Administration for Market Regulation, a "100 National Standards for AI" action plan is underway, coordinating 146 AI national standard projects. These cover frontier technologies including agents, embodied intelligence, and world models, as well as infrastructure standards for computing power, simulation platforms, and industry-specific large model applications in sectors such as steel, power, construction, and transportation.