At the World Artificial Intelligence Conference in Shanghai this July, China signaled a shift in its approach to the next phase of the AI revolution. While much of the global discourse remains fixated on the risks of large language models, Beijing has moved to define the architecture of the agentic era. The release of the Global Cooperation Initiative on Agent Mutual Trust, Interconnection, and Interoperability on July 17 is less a regulatory decree and more an attempt to establish the technical grammar for how autonomous systems will communicate across borders.
The speed of this development reveals a deliberate, synchronized strategy. In just 80 days, China moved from domestic policy to national standards, and finally to a global outreach framework. It began on May 8, when the Cyberspace Administration of China, alongside the National Development and Reform Commission and the Ministry of Industry and Information Technology, issued the first national guidelines for AI agents. This policy defined agents as systems capable of autonomous perception, memory, and execution, while setting a target for 70% adoption in smart terminals by 2027. By June 26, this policy was codified into technical reality with the release of GB/Z 185-2026, a set of national standards for agent interconnection developed by the China Electronics Standardization Institute and over 70 key enterprises.
This technical foundation is the core of the new initiative. GB/Z 185-2026 provides a blueprint for agent discovery, capability description, and interaction protocols across seven parts covering everything from identity codes to external tool invocation. By focusing on identity codes and mutual authentication, the standard aims to solve a looming problem: the creation of isolated, incompatible ecosystems. If OpenAI, Google, Microsoft, and Chinese firms continue to build agents using proprietary, non-communicative interfaces, the result will be a fragmented digital landscape. China’s initiative proposes a common language to prevent these islands of intelligence, positioning its own technical standards as the bridge.
The timing of this initiative contrasts with the current regulatory posture of the West. While China has spent the spring building an infrastructure layer, the United States and the European Union remain in a different phase of the governance cycle. As of early July, the US government has provided no specific federal guidance on agentic AI, with NIST frameworks not expected until 2027 or later. The Congressional Research Service confirmed this gap explicitly in its July 6 report (IF13151), noting no known US government guidance specifically addressing agentic AI. Meanwhile, the EU is preoccupied with the enforcement of the AI Act, which, according to the EU AI Office, treats agent-specific considerations as only preliminary. France’s antitrust focus on agent market concentration, issued the same day as the Chinese initiative, further underscores that Western policy is oriented toward competition and compliance rather than the technical interoperability of autonomous systems.
This gap creates a strategic opening for China to position itself as the primary convener of global AI governance. The WAIC 2026 forum, which included a High-Level Meeting on Global AI Governance, served as the venue for this positioning. By offering a framework that emphasizes joint infrastructure development and lower entry costs, China is explicitly targeting developing nations. The initiative’s focus on bridging the intelligence divide suggests an effort to export its technical standards to markets that have yet to lock into Western-centric agent ecosystems. For these nations, the promise of a pre-built, interoperable framework is a compelling alternative to navigating a fragmented global market.
The ten-point framework covers cooperation on foundational technologies including digital identity and mutual recognition, open international standards for agent interaction, joint AI infrastructure development, safe cross-border data flow respecting national security and privacy regulations, and security and ethics across the entire agent lifecycle. This is not abstract governance language — it maps directly to the technical specifications in GB/Z 185-2026, which defines the trusted identity cornerstone: identity codes, identity accounts, credentials, and mutual authentication for cross-system, multi-agent collaboration.
It is essential to recognize that this initiative is a cooperation framework, not a binding regulation. Its influence will not be determined by mandate, but by the extent to which international partners choose to adopt these standards. The real question is whether the technical specifications in GB/Z 185-2026 prove useful enough to become the de facto global standard for agent interaction. If they do, China will have established the infrastructure layer of the agentic economy while the US and EU were still focused on other questions.