판 이력 — Escaping the Agreement Trap: Defensibility Signals for Evaluating Rule-Governed AI | AIChainDay