Your AI runs on vibes, not rules: Can you prove your agent made the right decision?

In December 2026, two months after this conference, the Privacy Act's automated decision-making disclosure obligations come into force. If your AI agent operates autonomously, can you produce the reasoning chain? 

The governance framework exists as a PDF on a SharePoint site. The agent has never read it.

The governance approaches most organisations rely on were designed for a world where AI generated text. We now live in a world where AI takes actions. This session examines why first-generation approaches (human-in-the-loop, training-time alignment, policy-as-code) have been outgrown by autonomous systems acting at machine speed across permeable boundaries, and presents three primitives for runtime governance through a constitutional layer: autonomy boundary classification (observe, recommend, act), supervisory control from formal control theory, and decision provenance via causal directed acyclic graphs. Autonomous vehicle engineering arrived at the same principle. Demonstrated through worked scenarios with a governance readiness assessment toward December 2026.