Seeing AI Agents Is Not Enough. Security Teams Must Enforce What They Can Do

The article argues that AI agent security must move beyond visibility to enforcing least privilege. It highlights challenges like prompt filtering and the need for granular access controls.