A reader asked a newspaper's AI what the mayor promised developers. It invented a quote. The gate caught it before a reader saw it. This is the failure that hit Ask The Post AI, whose AI-generated podcast invented quotes under the paper's own masthead (Semafor, Dec 11 2025). The moment a newsroom answers from its own archive, it owns the words, and there is no Section 230 shield for what your system produces. The Standards Desk is the provenance gate we built to wrap a publisher's conversational engine. In our demo it runs over The Riverbend Ledger, a synthetic 13-year archive (everything here is fictional). A reader asks what Mayor Reyes promised developers about the riverfront parcel. The archive records no such promise (one article notes the parcel "was not on the agenda"). The answerer confabulates one anyway, the way any plain vector-RAG widget would. Here the decision is something you can see. Every claim is grounded to a real passage, every quoted string is checked verbatim against its source, and the fabricated quote fails that check. The banner flips to HELD FOR STANDARDS-DESK REVIEW, the offending span highlighted, the answer blocked from readers and routed to a human editor. Never shown to readers, never a bluff. Underneath, that verbatim-quote rule and the policy gate are plain Python, outside the model. Agents advise, code decides. Publish-safety never rests on a model's self-report, which is why a better model does not retire the gate. Whether an answer is safe to publish under your masthead is a governance property, not a model property. If your newsroom is working out where AI drafting ends and the standards desk has to begin, we would genuinely like to hear how you are drawing that line. 📰 #MediaTech #AIGovernance #NewsroomAI #EditorialStandards #PublisherTech