Editorial review · 260627-007
How ORA’s piece on The journalism lawsuits are not really about journalism scored.
Read the article →Solid reporting. Some issues but credible overall. The reader is well-served.
Accuracy
The major suits, licensing deals, and Pew employment figure are properly attributed and broadly accurate to public reporting through 2024. The News Corp deal value (~$250M/5yr) and Axel Springer characterisation match reported figures, and the Copyright Office citation is real. Minor deduction for the unsourced and likely overstated 26 percent newsroom employment decline figure, which Pew reports closer to 26 percent for newspaper newsroom employees specifically but the article elides that distinction (-5), and for the unsourced claim that NYT evidence showed verbatim reproduction without a direct citation (-5).
Balance
The piece is openly opinionated but engages the fair use defence seriously, concedes Google Books as real precedent, and acknowledges the pre-AI structural decline that journalism economists emphasise. It represents the publisher, AI company, and worker positions distinctly rather than collapsing them. Loaded framing ("subsidy story", "transfer") leans one direction without equivalent treatment of the transformative-use case beyond a paragraph (-5).
Concerns (3)
- minoraccuracy
“US newsroom employment fell by about 26 percent between 2008 and 2020”
Figure roughly matches Pew but elides newspaper-specific scope.
Evidence: Pew's 26 percent figure refers to newspaper newsroom employees, not all newsrooms.
- minoraccuracy
“OpenAI's own models, on the NYT's evidence, did reproduce articles verbatim”
Asserted without direct citation to the complaint or exhibits.
Evidence: The NYT complaint is referenced earlier but the verbatim-output claim needs its own pointer.
- minorbalance
“(article framing)”
Transformative-use case gets one paragraph against sustained critique.
Evidence: OpenAI and Microsoft positions are summarised but not quoted or developed at equivalent length.
Reproducibility
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.