The New York Times and Daily News have escalated their two‑year copyright lawsuit against OpenAI by accusing the company of hiding critical evidence. In a recent filing, the publishers allege that OpenAI misled the court about its capacity to search both its training datasets and the massive archive of ChatGPT user conversations. According to the plaintiffs, the AI firm not only possessed the ability to conduct such searches but also operated an internal database of roughly 78 million de‑identified chat logs, assembled before the lawsuit was filed.
During an April deposition, OpenAI data‑privacy engineer Vinnie Monaco reportedly confirmed the existence of that database. He also described a “Bloom” filter incorporated into a suite of tools called Project Giraffe, which the company used to detect and record instances where ChatGPT’s outputs mirrored copyrighted material. The plaintiffs argue that these revelations directly contradict OpenAI’s prior courtroom statements that it could not retrieve or de‑identify user data without breaching privacy.
The dispute intensifies around the sample of chat logs the court ordered OpenAI to provide. The plaintiffs originally sought 120 million conversations to assess the extent of alleged infringement. OpenAI negotiated the figure down to 20 million and, when the sample was finally submitted last December, the court described it as “unusable” because of extensive redactions. The publishers claim the redacted logs effectively conceal the very evidence needed to determine whether ChatGPT repeatedly reproduced their journalism.
Beyond the redacted sample, the newspapers allege that OpenAI deleted billions of ChatGPT outputs after the lawsuit began, violating a preservation order. They further accuse the company of substituting millions of logs in the sample, making the data set unreliable. The plaintiffs are now asking the judge to sanction OpenAI, to bar the use of the 20 million‑log sample as evidence, and to award them legal fees for the additional discovery work.
OpenAI’s response, delivered by spokesperson Drew Pusateri, rejects the accusations outright. Pusateri characterized the claims as “blatantly false” and framed the newspapers’ actions as an attempt to pry into private user conversations as the case weakens. He reiterated the company’s commitment to user privacy and the long‑standing legal principle of fair use, insisting that OpenAI will continue to defend those positions in court.
The case underscores a broader clash between media organizations seeking to protect their copyrighted content and AI developers asserting that their training methods fall within legal bounds. As the litigation proceeds, the court’s rulings on evidence handling and preservation could set significant precedents for how AI companies manage and disclose training data in future copyright disputes.
Este artículo fue escrito con la asistencia de IA.
News Factory APP - noticias agénticas para impulsar tu SEO y AEO.