On May 6, reports emerged indicating that, under the auspices of the China Institute of Communications, the China Academy of Information and Communications Technology, in partnership with prominent cloud service providers, has recently developed the 'Self-Regulatory Framework for Network and Data Security in Cloud-Based Intelligent Agent Services (2026 Edition)'. This framework is designed to set safety benchmarks and behavioral standards for the secure advancement of cloud-based intelligent agents, promoting a harmonious balance between high-quality development and robust security in AI innovation applications. Comprising four chapters and twelve articles, the framework is tailored for scenarios where cloud service providers supply pre-installed intelligent agent services, exemplified by OpenClaw. Centered around cloud service providers as the primary accountable parties, it outlines specific and actionable self-regulatory guidelines across six key domains: responsibility allocation, network security, data security, protection from malicious activities, user safety education, and risk information disclosure.