Check Point Research discovered how a single overlooked API in LangGraph, one of the world’s most widely used AI agent frameworks, can hand an attacker complete control of your AI infrastructure.