However, I also think the author should expand their definition of what constitutes "security" in the context of agentic AI.