AI agents use the same networking infrastructure as users and apps. So security solutions like zero trust should evolve to protect agentic AI communications.