The ServiceNow Australia release shifts AI governance from policy to enforcement — blocking unapproved agents at the infrastructure layer, not flagging them after the fact. This decision-grade synthesis covers what materially changes from Zurich, seven critical gaps that remain, and a phased roadmap structured by role for CISOs, architects, AI engineers, and TPMs. GA target: May 2026.