Sovereign Cyber AI: Beyond the obvious

Sovereign AI usually raises three concerns: where the data resides, what the regulator will ask, and a geopolitically induced coma. Those are the obvious ones. The harder questions are who can alter your AI, whether your defenders can live without it, and one rarely asked: how far an attacker can steer it without ever touching it. Drawing on real-world initiatives, this session sets out a reference architecture for Sovereign Cyber AI and its value-realisation model. Existing standards stop short of this, and closing the gap is our job as a community. The safeguards we settle on now are what the next generation inherits.