Overview
- The New York Times and roughly 15 other news organizations filed a motion asking a Manhattan federal court on July 9, 2026, to sanction OpenAI for alleged evidence concealment in their long‑running copyright suit.
- The filing accuses OpenAI of repeatedly misrepresenting that it could not search model training data or output logs while having performed internal searches and of deleting or making billions of ChatGPT conversations unsearchable.
- Plaintiffs say deposition testimony from OpenAI engineer Vinnie Monaco shows the company created a large de‑identified conversation database (reported at about 78 million entries) and built tooling called a Bloom filter or Project Giraffe to detect regurgitated outputs.
- The publishers ask the court to award attorneys’ fees, bar OpenAI from relying on the company’s provided 20‑million‑conversation sample, accept adverse factual findings about use of their work, and impose other sanctions for discovery misconduct.
- OpenAI denies the accusations, citing user privacy and its fair‑use defense, and the dispute remains active in SDNY with outcomes that could affect how commercial AI firms gather training data and how news publishers are paid or protected.