Edge AI in defence fails not because of the model but because of the operating model. This post examines what breaks in real deployments across land, maritime, and air environments and why it matters.
LLM-enabled adversaries change the cost structure of information operations permanently. This post makes the case for wargaming against AI-generated pressure and explains what you are actually testing for when you do it right.
Sovereign AI is not about where servers sit. It is about who controls the system when it matters. This post unpacks real sovereignty in defence procurement and why buying a postcode is not enough.
AI assurance is not confidence or compliance theatre. It is evidence. Defence needs proof of bounded behaviour, policy enforcement, and reconstructability under real conditions, with change control that sustains assurance after go live, not just at launch.