New York Times says OpenAI hid evidence in ChatGPT copyright trial

In a dramatic turn of events, The New York Times and The Daily News have accused OpenAI of misleading the court regarding its capabilities to access customer chat logs and training data in a contentious copyright lawsuit. This legal battle has been ongoing for two years, centering around allegations that OpenAI improperly trained its generative AI models on the Times’ content, ultimately reproducing that journalism in responses generated by its AI. OpenAI has consistently maintained that it lacks the tools necessary to search its own training data. The company has also expressed concerns that retrieving and processing its extensive ChatGPT conversation logs could pose significant technical challenges and infringe on user privacy. The news organizations involved have sought access to this data to ascertain whether their copyrighted materials were included in OpenAI's training datasets and how frequently ChatGPT might have utilized or replicated their content. During a court-ordered deposition in April, OpenAI’s data privacy engineer, Vinnie Monaco, reportedly disclosed that internal searches had indeed been conducted to identify copyrighted journalism within the training corpus. Furthermore, it was revealed that OpenAI had compiled a database of approximately 78 million de-identified ChatGPT interactions prior to the lawsuit, a move aimed at assessing its potential copyright infringements. Additionally, OpenAI allegedly introduced a tool known as the “Bloom” filter as part of a project dubbed “Project Giraffe” shortly after the lawsuit commenced. This tool was designed to detect and catalog instances of content reproduction in AI outputs. These revelations have raised significant eyebrows, particularly given that the plaintiffs originally requested a sample of 120 million chat logs, which OpenAI later negotiated down to 20 million. Although OpenAI submitted this reduced sample to the court in December, it was criticized for containing excessive redactions that rendered it nearly unusable. The plaintiffs have further claimed that OpenAI deleted billions of ChatGPT outputs following the initiation of the lawsuit, allegedly breaching a court order designed to preserve evidence. They assert that OpenAI manipulated the sample provided to the court, complicating their efforts to obtain previously collected information. Lead counsel Ian B. Crosby remarked, "If OpenAI genuinely believed that copying our clients’ journalism was fair and legal, it wouldn’t have hidden the truth about having done it." The New York Times and The Daily News are now urging the judge to impose sanctions on OpenAI for supposedly withholding crucial evidence and mishandling the discovery process. Their requests include preventing OpenAI from using the 20 million chat log sample as evidence, acknowledging that the logs would likely reveal substantial replication of the plaintiffs’ content, and compensating for legal costs incurred in the pursuit of this evidence. In response, OpenAI spokesperson Drew Pusateri rejected these allegations, asserting that the Times is attempting to invade the privacy of unrelated users as their case weakens. "As the Times’ case weakens and they’ve been forced to drop claims against us, they’re persisting with their efforts to invade the privacy of people who have nothing to do with this case, including by making these blatantly false allegations," Pusateri stated. He affirmed OpenAI's commitment to protecting user privacy and upholding the principles of fair use.