Overview
- Citing unsealed court filings released Thursday, publishers disclosed quotes from OpenAI and Microsoft executives who described models as “substitutive” of news and an “existential threat” to journalism.
- The filings say OpenAI’s mid‑training datasets contained at least 91,692 plaintiff works and that Common Crawl pulls included more than two million nytimes.com documents, which plaintiffs say shows mass copying for model training.
- Documents and deposition excerpts allege engineers discussed a “hack to get around” the New York Times paywall and described efforts to strip copyright notices before data reached models.
- Internal Microsoft metrics in the filings show its Copilot answer engine cut click‑throughs to the New York Times by as much as 93% and warn of a “doom loop” that could harm publishers and the web’s content ecosystem.
- The case is now before Judge Sidney H. Stein with summary judgment briefing underway and a DOJ brief supporting the tech companies’ fair‑use argument, and a ruling for publishers could force licensing deals or changes to how models are trained, with particular risk to local newsrooms.