Every AI Policy Ever · open in the interactive register
Published (non-binding, current)Non-binding · Guidelines / framework
网络安全标准实践指南——智能体部署使用安全指引
Record CN-26 · TC260-PG-20266A
TC260 practice guide on the security of deploying and using agents: large-model-based AI agents for personal-assistant use that need high user permissions. For the assessment, preparation, deployment, use and decommissioning stages it advises against unmaintained projects and agents that open public interfaces automatically, and calls for installation materials from official channels, a separate device or isolated environment, large models that have completed generative AI service filing, no administrator rights, limited directories and network exposure, full logging, a high-risk operations list with second confirmation or blocking, and emergency stop.
No English version of this instrument was found. The English title and the summary are the register's own, written from the sources below; for the exact wording use the Chinese text.
Version v1.0-202607, released with notice 网安秘字〔2026〕80号 of 1 July 2026. It can also guide the choice of commercial agent services. High-risk operations include stopping system services, formatting disks, bulk deletion, changing permissions, keys or firewall rules and accessing password managers, and may include transfers and payments. Annex A is a security checklist; Annex B gives organisations management guidance: an internal agent-use policy with prohibited behaviour and approval steps, an agent asset register, logging and risk analysis of agent activity, and discovery of unapproved agents on internal networks.
Open the official source https://www.tc260.org.cn/tc260/sjzn/202608/67702dcf57214da792499be8518446dc/files/f3be1cc4d35d428c98a9f5d197c963ac.pdf
Secondary source: https://www.tc260.org.cn/tc260/tzgg/202607/98b6f259f8a3428b8fe1cfea3c0fb19e.shtml
Archived. The secondary source is a bot-check or empty page (The page has very little readable text (599 characters); it probably needs JavaScript). Last attempt 2026-10-18 00:45:31 UTC
Warning: The page has very little readable text (599 characters); it probably needs JavaScript
TC260-PG-20267A (security guide for individual AI users) China · Published (non-binding, current)
TC260 draft guide on agentic AI system development security China · Draft / proposed
National Cybersecurity Standardization Technical Committee (TC260) Secretariat. Cybersecurity Standard Practice Guide — Security Guidelines for the Deployment and Use of Agents [TC260-PG-20266A]. 2026-07-01. https://www.tc260.org.cn/tc260/sjzn/202608/67702dcf57214da792499be8518446dc/files/f3be1cc4d35d428c98a9f5d197c963ac.pdf. Status: Published (non-binding, current); verified 2026-10-18 (level B). Local copy archive/_runs/2026-10-14-discover-leads-2026-10-14/C010/2026-10-14/official.pdf retrieved 2026-10-14 13:10:56 UTC, SHA-256 11124dbe739eebd438a6ed5e9b612fae599aa116d74ceab05a9d1c426f101ec4. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record CN-26.
Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.