Tags发布时间Nov 19, 2025摘要Researcher shows how agentic AI is vulnerable to hijacking to subvert an agents goals and how agent interaction can be altered to compromise whole networks.来源Dark Reading