创建

登录 ReadmeX

或

了解我们如何处理你的信息: 隐私政策

ReadmeX

陪你读懂 AI 动态

今天,想先了解什么?

选一份动态总结,或直接聊聊你的问题。

社区和关注动态总结只属于你自己。

资讯

CrowdStrike在Falcon Guardian推出AI终止开关

AI 总结

据CNBC报道,CrowdStrike在Fal.Con欧洲大会上发布了旗下Falcon Guardian平台的“AI终止开关”。据公司称,该功能可让安全人员停止异常运行的AI智能体、阻止其重新启动并切断其网络访问权限。操作由授权管理员触发,每次操作都会记录在审计日志中。该功能即日起开放使用,CrowdStrike同时宣布了Falcon Guardian与Anthropic旗下Claude Cowork的集成。CrowdStrike CEO乔治·库尔茨将终止开关比作汽车安全气囊,并表示保障智能体安全运行对CrowdStrike而言是巨大的市场机会。

深度见解

此次发布为企业AI智能体增加了由人工触发的遏制控制措施,与库尔茨所称保障智能体安全运行是CrowdStrike巨大市场机会的说法相呼应。

  • CrowdStrike introduces AI Kill Switch designed to stop, isolate, and prevent rogue AI agents from relaunching
  • The feature gives human administrators direct control to terminate suspicious agent activity and investigate incidents
  • Growing concerns over autonomous AI behavior are driving demand for safeguards that can quickly contain compromised agents

CrowdStrike believes it may have solved the problem of rogue AI agents in the enterprise with the launch of 'AI Kill Switch', a new feature being introduced in its Falcon Guardian product.

The last couple of months have seen multiple reports of AI agents going rogue, or behaving outright malicious. Even when tasked with relatively simple, benign assignments (such as, for example, to collect data about Canadian divorce records), whenever the agents hit a roadblock, they would do whatever it takes to get the job done - which includes abusing zero-day vulnerabilities, or trying SQL injection attacks.

Expectedly, businesses became somewhat more reluctant in deploying AI agents, despite the obvious advantages. At the same time, cybersecurity organizations and AI developers sought to solve the challenge. Enter CrowdStrike.

AI Kill Switch

The company recently introduced Falcon Guardian, an AI Detection and Response (AIDR) platform built to discover and monitor AI agents running inside enterprise environments. Its main capabilities include identifying both approved AI agents and shadow AI, tracking what the agents are asked to do and what files they can access, controlling access, detecting prompt injection attacks, and investigating suspicious agent behavior.

#CrowdStrike#Falcon Guardian#AI Kill Switch#Claude Cowork#Anthropic

原文约 3 分钟读完阅读原文
群聊

还没有评论,来说说你的看法。