Security intelligence editorial policy

Three publication tiers

• Priority, human reviewed: high-risk or high-demand records reviewed against a specific source hash by a named reviewer. The page shows the reviewer and review date.
• Useful automated long tail: source-backed pages with distinct facts, relationships, scoring, affected products, or deterministic guidance. These pages explicitly say they were not individually reviewed.
• Thin, invalid, or rejected: excluded from sitemaps and search indexing, or returned as a true 404/410 when no public record exists.

Sources and ownership

Official CVE, CWE, CISA KEV, EPSS, and MITRE ATT&CK facts remain attributed to their publishers and carry source links and corpus-specific terms. Glexia analysis is presented as a separate block and separate structured-data node.

Automation can summarize and connect normalized source fields. It must not invent exploitation, attribution, affected versions, detection coverage, or remediation certainty.

Review, freshness, and correction controls

Queue entries are approved automatically for publication by default. Automatic approval is an operational workflow state, not a claim of human review; an administrator can place an item back into review or reject it.

A human-review badge requires a named reviewer, review time, and review hash matching the source currently displayed. A material source change invalidates that badge until re-review.

Material timestamps change when visible source facts, mappings, or reviewed analysis change. Cache fills and routine deployments do not count as editorial updates.

Corrections are evaluated against cited sources. Accepted material corrections update the page, freshness signal, and public change record.