A Methodology for Auditable Trustworthiness Levels in AI Lifecycle Governance

This paper proposes a lightweight methodology for auditable trustworthiness levels in AI lifecycle governance. It combines a formal, context-sensitive representation of trustworthiness profiles with a lifecycle procedure for labeling, monitoring, reassessment, and reporting. Decision trees serve as an interpretable proof-of-concept for learning explicit trustworthiness plateaus and level transitions. The methodology also introduces boundary margins and profile drift as lifecycle diagnostics. The authors illustrate the approach with synthetic lifecycle traces covering degradation, shocks, updates, heterogeneous monitoring cadences, and system comparison. It is intended to support conformity documentation, not replace legal or expert judgment.