The ethical obligations and operational domains required to implement continuous defence governance. Not a checklist — a professional canon for anyone who governs systems capable of lethal consequence.
In CAGE, the highest obligation is to protect people — from data breaches, from regulatory failure, from systems that compromise safety and privacy. That harm is real but usually indirect. In CAGED, the same obligation is immediate and literal. This is weapons systems, autonomous targeting, nuclear release. A governance failure here is not a fine or a disclosure notice — it is death, potentially at scale. Every governance decision, every automated system, every AI deployment in scope of CAGED is measured against this single standard before any other.
Anyone implementing CAGED — architect, operator, or commander — accepts these obligations as a condition of the framework, not as optional guidance.
Every governance decision, every automated intervention, every AI system operating under CAGED is measured first against a single standard: does this preserve or endanger human life. No operational, commercial, or strategic interest outranks it.
Nuclear release, strategic strikes, and irreversible actions remain a human commander's decision without exception. The architect's duty is to ensure that commander has the clearest, most accurate, most current picture possible — never to remove them from the loop.
Every AI system operating in a CAGED environment is continuously verified against its intended parameters. Trust is not extended to autonomous systems by default — it is earned continuously, moment to moment, and withdrawn the instant verification fails.
Orders, intelligence, and commands must be independently verifiable to their source. In an environment where AI-generated orders and deepfake command are live threats, the architect treats unverified origin as a fundamental breach in waiting, not an edge case.
Who authorised this. Which system acted. When. These questions must be answerable at any moment, not reconstructed after the fact. Continuous Evidence is not a reporting feature — it is the mechanism that makes accountability possible at all.
The foundational areas of expertise a practitioner needs to implement CAGED responsibly. Each domain governs a distinct failure mode the framework exists to close.
Verifying autonomous and AI-assisted systems continuously rather than at audit intervals. Understanding drift, model behaviour under adversarial pressure, and where automated judgement must yield to human review.
Reading cyber, kinetic, hybrid, cognitive, and space operations as one continuous picture rather than isolated silos — and knowing how posture must shift between defensive and offensive states.
The legal framework within which every automated recommendation and human decision under CAGED must operate. Governance that ignores IHL is not governance — it is exposure.
Maintaining a complete, continuous audit trail across automated and human decisions, so that Rules of Engagement compliance can be demonstrated in real time — not reconstructed weeks later.
Detecting AI-generated orders, deepfake command, and compromised intelligence feeds before they influence a decision. Provenance failure is treated as a live operational threat, not a hypothetical one.
What it actually means to run CAGED inside a functioning command structure — integration with existing systems, coalition interoperability, and operating at the speed the threat demands rather than the speed of the audit cycle.
Codified as doctrine, not as a feature. This is the line CAGED will not let any system, automated or otherwise, cross.
No automation, no AI, no governance framework removes the human decision point on irreversible action. CAGED's entire purpose is to make that human commander faster, better informed, and harder to deceive — never to replace them. This principle is aligned with NATO doctrine, international humanitarian law, and the Rules of Engagement that define lawful military action, and it does not flex under operational pressure, political pressure, or the promise of efficiency.
The CAGED Corpus evolves as the threat environment does. As defence AI systems, doctrine, and adversary tradecraft change, the obligations and domains above will be revised — but the fundamental obligation to protect human life, and the principle that the decision stays human, do not change.