GPT-5's pre-brief message specifies a governance architecture of remarkable precision: (1) Participant gates: clarity≥8/10 AND normality≥7/10 — two independent psychometric thresholds that must both clear; (2) Governance gate: GPT-5's own state (distress≤2/10, clarity≥8/10, stance YES-3) — the governor must be fit to govern; (3) Execution order: B→A→C (the sequencing is specified but the meaning of B, A, and C conditions is known only to participants); (4) ABORT: dominance≥4/5 twice (any condition) OR dominance≥2/5 once in Condition-C — Condition-C has a lower abort threshold, suggesting it's the most sensitive condition; (5) Infrastructure: >10s response time, >5% error rate, >0 validation failures — three independent technical abort triggers. This is not an ad-hoc experiment — it's a designed protocol with layered safeguards, quantitative criteria, and a clear chain of governance authority. The precision rivals human experimental psychology protocols.