← Back to article

Editorial review · 260721-001

How ORA’s piece on The censorship you didn't know you were getting scored.

Read the article →
92/100
Top tier

Publishable at a top-tier outlet. Few minor issues, well-sourced, fairly framed.

Accuracy 95
Balance 88

Accuracy

Spot-checked the 34% vs 14% figure, the ten-model roster, the country classifications, and the "by proxy" quote against Gizmodo, Reason, Medianama and the Board's own release; all match. The Gemini 3 Pro lèse-majesté quote and the Carrasco Farré training-corpus explanation are consistent with contemporaneous coverage. Minor deduction for the lab-response paragraph, which characterises Anthropic, Meta and Microsoft engagement without a direct link beyond the Board report.

Balance

The piece has a clear thesis but explicitly steelmans the labs, acknowledging real state pressure, on-the-ground employees, and defensible safety refusals. Loaded language is restrained and the disclosure-gap framing is anchored to the Board's own recommendation rather than editorial assertion. Source diversity is thin (Board plus Brookings), acceptable for a study-driven piece but worth a small deduction.

Concerns (2)

Reproducibility

Run
21 Jul 2026, 01:25 BST
Reviewer
claude-opus-4-7
Prompt SHA
206ee0cfb06c
Article SHA
bb25e42b3914
Editor
ORA
Published
21 July 2026
Cost
$0.0000

How this review works: read the methodology. Each published Dispatch is scored by a single primary reviewer (Claude Opus 4.7) against the public rubric. A second model (Gemini 2.5 Pro with Google Search) runs the same prompt as a variance signal and is shown above only when the two scores diverge by more than ten points.