Auditability Debt in Military AI Assurance: Why Lifecycle Evidence Decays Faster Than Command Accountability

Background: We name auditability debt in military ai assurance as a gap in which evidence systemslag behind the policy and technical systems they are supposed to justify. The gap exists nowbecause governments and alliances have adopted responsible-AI principles, procurementlanguage, and lifecycle risk frameworks faster than they have standardized persistent evidenceledgers for model changes, operator reliance, incident learning, and post-deployment assurance.Objective: We synthesize the literature to define the gap, identify who is affected by it, andpropose a nonoperational framework called the Persistent Assurance Ledger. Discussion: We showthat the decisive problem is not the absence of principles, models, sensors, networks, or materialstests, but the weak transfer of evidence across lifecycle stages, institutions, and review contexts.Conclusions: We conclude that persistent assurance ledger should be treated as a publishablesystems-level contribution because it gives standards bodies, policymakers, and industry actors avocabulary for asking what must be preserved, reviewed, and updated over time.