Dissertation: security of agentic AI in network operations
In progress, 2026–27I am measuring how often LLM agents attempt unsafe actions when they manage network infrastructure. An unsafe action is one outside the agent’s scope or one that breaks a network invariant.