en.Wedoany.com Reported - On May 7, 2026, the China Academy of Information and Communications Technology (CAICT) officially announced the establishment of the "Agent Open Source Community," backed by the Cloud Computing Open Source Industry Alliance and jointly initiated with multiple core industry players. Simultaneously, it released version 1.0 of the Agent Trust Handshake Protocol (ATH), which has been officially open-sourced on both the AtomGit and GitHub platforms.
The background for CAICT initiating this community lies in the severe security and trust challenges brought about by the rapid proliferation of open-source agent technologies in cross-platform, cross-entity collaboration scenarios. Currently, agent interactions face core bottlenecks such as the lack of identity authentication mechanisms, insufficient authorization and permission controls, non-traceable interaction behaviors, and difficulty in safeguarding user data sovereignty. Therefore, the newly established community aims to shore up the security weaknesses in industrial interactions, break down trust barriers between different open-source agent ecosystems, and build an open, standardized, and secure industrial development environment. The community will focus on building the open-source agent ecosystem and constructing a trusted interaction standard system, bringing together enterprises, research institutions, and other forces to fully promote the secure iteration and large-scale implementation of open-source agent technologies.
The ATH 1.0 protocol is precisely the technical standard designed to address these pain points. Adhering to six core design principles—"User Sovereignty, Three-Party Participation, Trusted Handshake, Decentralization, Least Privilege, and Full Traceability"—it establishes a professional, universal, and standardized trusted interaction framework for multi-agent, cross-entity, and cross-platform collaborative scenarios. Its core lies in accurately covering key security scenarios such as AI agent identity mutual verification, access permission control, and behavior auditing, providing clear technical standards and implementation guidelines for agent interaction activities across the entire industry.
From a technical architecture perspective, the ATH protocol constructs a dedicated handshake mechanism involving three parties: "User, Agent, and Application." Its core is a 9-step trusted handshake process comprising two independent phases, designed to complete decentralized communication authentication and establish a secure session. The protocol plays a unique role within the technical system, positioned as a "security checkpoint" before an agent's external interactions. Compared to mainstream agent communication protocols in the industry, such as MCP (Model Context Protocol) and A2A, ATH does not directly participate in communication. Instead, it builds an independent trust layer between the application layer and the communication protocol layer. By implementing fine-grained permission control and full-link behavior tracing through "authentication before communication" on top of protocols like MCP/A2A, it compensates for the deficiencies of existing protocols in permission granularity and security systems. This complementary and enhancing design allows it to be widely adapted to agent interaction scenarios across various fields, including enterprise office, lifestyle services, and industrial systems.
The release of this protocol brought together a powerful industry lineup. Presided over by Xu Zhiyuan, Deputy Chief Engineer of CAICT, and Li Wei, Deputy Director of the Cloud Computing and Big Data Research Institute, representatives from China Telecom Corporation Limited, China United Network Communications Group Co., Ltd., China Mobile Jiutian Artificial Intelligence Technology (Beijing) Co., Ltd., Tencent Technology (Beijing) Co., Ltd., Huawei Technologies Co., Ltd., ZTE Corporation, and The Chinese University of Hong Kong, Shenzhen jointly completed the launch of ATH 1.0. This action marks a crucial step for the industry in jointly building trusted interaction infrastructure for agents.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
