嗯，用户让我帮忙总结一篇文章，控制在一百字以内，而且不需要特定的开头。首先，我得仔细阅读文章内容，抓住主要观点。 文章讲的是自动化安全从脆弱的脚本转向自主代理，提到了“Governed Agency”框架，基于复杂系统理论。关键点包括TOTE循环、认知光锥和生物电信测量。这些都是为了防止自主代理追求局部目标而损害整体安全。 接下来，我需要把这些要点浓缩成一句话，确保不超过一百字。同时，语言要简洁明了，直接描述文章内容。 最后，检查一下是否符合用户的要求，没有使用“文章总结”之类的开头，并且控制在字数范围内。 </think> 文章探讨了从脆弱脚本转向自主安全代理的风险与解决方案,提出"Governed Agency"框架,通过定义目标状态和限制代理感知范围,避免局部优化损害整体安全。 2026-3-18 04:31:32 Author: www.reddit.com(查看原文) 阅读量:5 收藏
Most of our automation is "Level 0"—fragile, linear scripts that break the moment a schema changes. As we move toward autonomous bots, we’re risking what I call "Security Cancer"—where agents pursue local goals at the expense of the company.
At 90PB/day scale, we implemented a strategic security engineering framework for Governed Agency. It’s based on complex systems theories. Instead of writing better scripts, we define the "Target Anatomy" (the healthy state of the network) and let agents innovate to reach it.
Key takeaways:
-
TOTE Loops: A new way to log for compliance.
-
Cognitive Light Cones: Managing blast radius by limiting what a bot can "see" and "think."
-
Bioelectric Telemetry: Using APIs as the shared reality for all security tools.
Check it out here: https://www.securesql.info/2026/01/31/season2-zeronoisecollective/
How are you all managing "Rogue Agency" in your automation today? Are you still managing scripts, or have you started managing "Goals"?
如有侵权请联系:admin#unsafe.sh