policy

AI Editorial Policy

Authority boundaries, evidence, author approval, model traceability, corrections, and appeals.

Version: 2026-08-30
Canonical URI: /legal/ai-editorial-policy
Fingerprint: SHA-256 43c9a55e5737720f523331c9ed2abded420775d8fca61b7fac0a3de298793110

Permitted actions

Authorized automation may classify, identify defects, normalize metadata, recommend prices and categories, propose accessibility and editorial changes, request evidence or revision, and approve qualifying publications under versioned policies.

Expression boundary

A substantive change to author expression requires author approval. The portal must show the issue, evidence, current and proposed content, reason, severity, and response.

Reproducibility

Consequential decisions retain model and version, policy and prompt contract version, extracted evidence, score or confidence where applicable, decision, action, actor, and correction or appeal history.