Your AI Agents Are Leaking Enterprise Data. Here's Why Nobody Is Watching.

The post states that many enterprise security teams focus on monitoring employee interactions with tools like ChatGPT, while AI agents autonomously query internal systems (e.g., CRM), chain multiple tool calls, and send accumulated sensitive data to external LLMs without human oversight. It highlights that in March 2026, critical vulnerabilities (CVE scores up to 9.3) were disclosed in LangChain and LangGraph, with 9 million LangGraph downloads occurring in a single week. The post also notes that existing security tools have a structural blind spot for this type of threat.