Editorial review · 260527-004
How ORA’s piece on The states are regulating healthcare AI because the harm was already on the record scored.
Read the article →Solid reporting. Some issues but credible overall. The reader is well-served.
Accuracy
Core claims about California SB 1120, the UnitedHealth nH Predict litigation, and the Cigna ProPublica reporting align with the public record (-0). Footnote 1's URL points to an EU AI Act compliance page, not a state-level healthcare AI tracker, which is a mis-citation on a load-bearing source (-5). Footnote 6 is self-flagged as unverified for a specific 18 percent figure, and the California 'less than three percent' denial-rate movement lacks a direct DMHC citation (-10).
Balance
The piece has a clear point of view but engages the Chamber's preemption argument and the contrarian case on encoded technology assumptions in their own terms rather than as strawmen. Loaded framings (compliance theatre, victims had lawyers) are present but matched by acknowledgement that algorithmic consistency produces measurable benefits. Source set leans on payer-critical voices (ProPublica, AMA, plaintiff filings) with no insurer or vendor response quoted (-8).
Concerns (4)
- minoraccuracy
“Holland & Knight, 'States Continue Efforts to Regulate AI in Healthcare,' April 2026”
Footnote URL resolves to an unrelated EU AI Act compliance article.
Evidence: The hklaw.com link in footnote 1 is about EU August 2026 deadlines, not state healthcare AI.
- majoraccuracy
“denial rates have moved by less than three percent”
Specific outcome figure attributed to DMHC reporting without a direct citation.
Evidence: Footnote 2 references Q1 2026 DMHC reporting generally with no link or document identifier.
- minoraccuracy
“reduced geographic variation in denial rates by 18 percent”
Author flags the JAMA citation as 'pending direct verification' in footnote 6.
Evidence: Footnote 6 admits the specific citation is unverified, yet the figure is used to anchor the contrarian case.
- minorbalance
“(source set)”
No insurer or AI vendor voice represented on a contested policy topic.
Evidence: Cited voices are plaintiffs, ProPublica, AMA, and the Chamber; no UnitedHealth or Cigna response included.
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.