A federal judge has mandated that OpenAI must produce more than 20 million anonymized ChatGPT chat logs in connection with a significant copyright infringement case involving The New York Times. The ruling, issued by U.S. Magistrate Judge Ona Wang, determined that the chat logs are pertinent to the claims made by the newspaper against OpenAI.
On May 13, 2025, Judge Wang instructed OpenAI to preserve all output log data that would typically be deleted as the case progressed. The court subsequently convened multiple settlement conferences with the involved parties regarding concerns over OpenAI’s deletion of consumer ChatGPT logs. This issue arose after The New York Times expressed alarm over the company’s actions in light of their ongoing lawsuit.
The newspaper requested that the court require OpenAI to provide a “statistically valid monthly sample” of its ChatGPT output logs from December 2022 through November 2024. The objective was to substantiate claims that OpenAI had unlawfully utilized content from The New York Times to train its artificial intelligence models. OpenAI initially resisted this request, citing the extensive time required to comply and later raised privacy concerns regarding the logs.
Following OpenAI’s refusal to provide the logs, The New York Times filed a motion to compel OpenAI to comply. Both parties presented their arguments both orally and in writing. On November 7, 2025, Judge Wang ruled in favor of the newspaper, ordering OpenAI to produce the requested logs.
Despite the ruling, OpenAI sought to stay the judge’s decision, calling for reconsideration of its motion. Judge Wang rejected this request, noting, “OpenAI’s motion for reconsideration should be denied because OpenAI has failed to ‘put forth any facts or law that the Court did not consider and that would compel a different conclusion.’”
Judge Wang further emphasized that the production of the 20 million ChatGPT logs is not only relevant to the case but also proportional to the needs of the proceedings. She pointed out that the total number of retained consumer output logs at OpenAI is in the tens of billions, making the requested sample a mere fraction of the total.
With the court’s order in place, OpenAI now has seven days to comply and produce the logs as directed. This case marks a critical moment in the intersection of artificial intelligence and intellectual property rights, raising significant questions about the use of copyrighted material in training AI systems.