Editorial standards

We have a point of view. These are the rules that keep it honest.

1. Our headline, their story. Every item is attributed and linked to its source. We summarize; we don't reproduce.
2. Reporting and opinion live in different places. The body of an item is what happened. Our take sits under Why it matters, labeled as ours. A reader should never have to guess which is which.
3. The take has to follow from the facts above it. No take on an item asserts a fact the reporting doesn't support.
4. True dates, or flagged ones. Old news runs marked catch-up or doesn't run. A date we can't confirm is marked date uncertain.
5. Single-source stories are flagged. Until a second source confirms them, they carry single-source.
6. PR is labeled PR, and industry money is named. A company announcing its own safety work, sustainability numbers or impact study is a primary source for what it claims, not proof that the claim is true. Industry-funded research is labeled.
7. Advocates get the same scrutiny as companies. Being on our side doesn't make a number right. An advocacy group's figure is attributed to the group, the same as a company's.
8. We don't manufacture outrage. No item says harm happened where the source says harm might happen. Projections are labeled projections; allegations are labeled allegations.
9. Corrections are appended and dated, never silent. Every one is listed on the corrections page.
10. Tips are credited by name or handle, every time, unless the tipster asks otherwise — and we protect anyone who asks.
11. No coverage for sale. Keeping AI Honest takes no money from companies it covers, and its publisher's other businesses get no coverage here.
12. The gate passes more than it publishes. Volume is the machine's job. Judgment is ours.

How the wire is made

A poller reads a watchlist of primary sources and reputable outlets — regulators, courts, research groups, advocacy organizations, environmental and tech-policy press — and queues anything on the beat. A desk working under a written charter checks each one against the rules above, rewrites the ones that clear the bar, and passes the rest with a reason. Drafting is AI-assisted; the editor is accountable for everything that runs. The sources we read.