AI Agent已进入实战攻击链,会话历史与配置文件还留在公开目录
ai_agentllm事件报告
RT @_josehelps: CrowdStrike just published a look at “agentic AI” used in real intrusions.
A likely Chinese-speaking, financially motivated actor hit South Korean financial orgs with ARTEX (an open-source agentic pentest suite) plus Claude Code. Open directories on their infra held Claude session histories, ARTEX configs, and Claude memory files.
https://t.co/nFPoKa8PLU