Umpiring AI decisions: The Decision Review System (DRS) framework for contestable and board-ready security governance

Your AI-powered SIEM has flagged a senior executive as "high risk" and automatically revoked her building access. She escalates the matter to the CEO, demanding an explanation. Your logs show "Model confidence: 68%," but offer no seed value, no prompt chain, and no reproducible decision path. This is not audit-ready, and it fails the "contestability" requirements of the Australian AI Safety Standard.

As both a cyber security leader and a NSWCUSA grade cricket umpire, I make over 300 split-second decisions per match. In both domains, pattern recognition under pressure is critical, and every decision must be logged, contestable, and defensible post-match. This session translates the Decision Review System into an enterprise AI governance model.

Drawing on decision frameworks from both elite cricket officiating and 21 years of securing enterprise infrastructure across three continents, we examine how to transform probabilistic AI outputs into board-defensible evidence. We will move past broad ethics to the technical implementation of "Contestability," ensuring your organisation is prepared for dual ISO 27001 and emerging ISO 42001 (AI Management System) certification. 

The session will cover:

• The DRS Technical Stack: Essential logging standards (temperature, seed values, and prompt templates) required for reproducible AI audits.
• Managing Probabilistic Risk: How to lead a Board through a risk landscape that has no fixed binary boundaries.
• The Accountability Gap: Defining ownership models for AI-assisted decisions to ensure compliance with sovereign safety mandates.

Attendees will explore a realistic tabletop scenario illustrating how DRS methods improve incident review and remediation workflows. By the end of the session, security leaders will have an actionable framework to enhance AI decision governance and strengthen their compliance posture in rapidly evolving environments.