Entities Launch AI Agent Self-Discipline Convention

People's Daily reported that the AI Agent Development Self-Discipline Convention was officially released during the World Internet Conference Wuzhen Summit. Co-initiated by 20 entities, including the China Academy of Information and Communications Technology, Huawei, Tencent, Ant Group, Alibaba, and Douyin, the document puts forward eight self-discipline measures. These measures focus on trust systems, interconnection infrastructure, interoperability, security protection, governance mechanisms, user rights, and industrial applications. The convention proposes advancing technologies and standards for identity registration, capability discovery, intent understanding, network routing, resource scheduling, and protocol compatibility. These efforts aim to lower cross-domain discovery and connection costs to support agent interconnection. On security protection, the convention calls for strengthening technical systems covering forward-looking risk assessment, operation monitoring, behavior management, and anomaly fusing. It also advocates joint risk information sharing to prevent unauthorized operations, privacy leaks, and out-of-control actions. Additionally, the convention encourages developing low-barrier development and deployment tools to embed agents into existing production systems, while deepening interface adaptation and respecting intellectual property rights to reduce interoperability barriers.