Review Queues And Lines Of Defence

Queue Taxonomy

Planned queues cover AI output concerns, source and provenance concerns, editorial framing concerns, wrong attribution, correction requests, appeal or reconsideration requests, transparency-log review, source-ingestion or data-quality issues, prompt/model/pipeline change review, incident review, legal-sensitive escalation, and future user-conduct moderation.

Each queue should identify intake source, likely content classes, first reviewer role, evidence required, possible outcomes, escalation triggers, and transparency-log consideration.

First Line Of Defence

AI and system triage should catch missing provenance, broken source mappings, missing generation records, failed prompt/model outputs, duplicate reports, and publication blockers. It should route reports by category and severity.

Second Line Of Defence

Fact-check or editorial review should compare output with source material, check attribution and omitted context, apply the Editorial Integrity Code, and decide whether to correct, regenerate, rewrite, remove, escalate, or record no action.

Third Line Of Defence

Safety lead, trained moderator, escalation reviewer, legal reviewer, or senior decision-maker should handle high-risk, contested, repeat, legal-sensitive, public-figure, incident, and future user-conduct appeal cases.

Evidence Record

Review records should connect report ID, content ID, content class, source object ID, generation ID where applicable, source snapshot or URL, output snapshot, issue category, severity, reviewer role, decision, rationale, action taken, escalation status, transparency-log decision, and restricted evidence references.