Lifecycle Governance Doctrine (LGD): Evidence-Gated AI Lifecycle Governance

Abstract

English formal term: Evidence-Gated AI Lifecycle Governance. 中文主权词: 凡自治之物 —— 有籍 · 有证 · 有门禁。 The Lifecycle Governance Doctrine (LGD) holds that governance of any autonomous system — AI, robots, autonomous vehicles, low-altitude systems, data assets — must be a full-lifecycle chain from birth to retirement, not single-point checks. It rests on three laws: Registry at birth (identity, creator, governance ownership), Evidence for every key act (logs, risk register, change records), and Gates before every evolution (trigger → assessment → authorization → review; humans hold sign-off). The reference model is medical-device lifecycle regulation, generalized across 12 domains. Positioning: aligned with GB/Z 185—2026 (China, AI agent interconnection), ISO/IEC 42001, NIST AI RMF, the EU AI Act and IETF/CSA directions. LGD contributes the implementation layer (How), a medical-device domain instantiation, and a cross-domain master template — it does not claim precedence over existing standards or prior work.