Editorial Standards
HORIZON publishes one type of content: structured outbreak intelligence with auditable source provenance. Every claim is traceable. Every correction is logged. This page documents the rules we hold ourselves to.
1. Source verification
Every record is tied to a named, published source. We do not accept anonymous tips. We do not republish unsourced social media posts as fact. Every fetched document is hashed (SHA-256) at ingestion under the Berkeley Protocol; the hash is stored alongside the fetch timestamp, URL, and User-Agent.
2. NATO Admiralty Scale rating
Every source is rated on reliability (A–F) and credibility (1–6) per NATO AJP-2.1. Auto-application of an extracted fact to the incident ontology requires an A1/A2/B1/B2 source or three corroborating independent sources within 48 hours. See the full procedure on the methodology page.
3. Dual confidence model
Pipeline confidence (machine, statistical) and analyst confidence (human, vetted) are stored as separate columns and never conflated. Front-end displays distinguish them visually. Export endpoints require analyst confidence on every record.
4. Corrections
When we discover an error, we (a) correct the record, (b) log the correction in our corrections log, and (c) preserve the prior version's hash in the database for audit. We never silently overwrite history.
5. No clinical advice
We are not a clinical service. We do not diagnose. We do not recommend treatment. Every medical topic page carries an explicit disclaimer directing readers to WHO, CDC, ECDC, or their local clinician.
6. Plagiarism + copyright
All prose on HORIZON is original. We cite WHO/CDC/ECDC/PAHO sources by name and link to them. We do not copy excerpts longer than a few words without attribution. Our open-data licence is CC BY 4.0; we ask the same attribution standard of anyone using our data.
7. AI training disclosure
HORIZON content is generated by humans. We use machine learning for
extraction and deduplication (specifically the qualification pipeline
described in our methodology). We do not generate prose with large
language models. Where ML touches the dataset, the
pipeline_confidence field is amber — not green.
8. Editorial independence
HORIZON receives no advertising revenue. We accept no sponsored content. We do not promote products, treatments, or commercial services.