How we decide what gets published.
The Archive only works if you can trust the floor under it. These are the rules that govern what we publish, how we label it, and how mistakes get corrected.
No guilt by association
Appearance in a flight log, a phone book, or a guest list is a documented connection — not a verdict. We never imply criminal conduct that has not been proven in court, and we label every record by what it actually shows.
Source-or-spike
If a claim has no working public source, it does not get published. Every assertion in a public record on The Archive carries at least one verifiable citation. Submissions without sources are returned, not posted.
AI as assistant, not author
We use language models to summarise, structure, and cross-check material. AI never originates a claim. Every fact is traced back to a human-verifiable document before it is shown.
Correctable by design
Records are versioned. If we publish something inaccurate, we want to know. Corrections are accepted via the submission form and the record is updated — or removed — when the underlying source is shown to be wrong.
Every claim is labeled, not levelled.
A rumor and a court filing are not the same thing — and we refuse to flatten them. Each record carries one of four tiers so you always know what kind of evidence you are looking at.
Backed by a primary document or court record. Cited directly.
Reported by an established outlet with editorial oversight. Cited inline.
Sworn testimony, court allegations, or credible witness claim. Clearly labeled as allegation.
Surfaced but not yet corroborated. Held in editorial review — not published publicly.
If a claim cannot survive being clicked through to its source, it does not belong on The Archive. That is the bar.