OpenAI must turn over 20 million anonymized ChatGPT conversation logs in a copyright fight with The New York Times and others. Chat logs are a critical piece of evidence in the investigation into whether ChatGPT is reproducing copyrighted news content without authorization, according to U.S. Magistrate Judge Ona Wang.
The order is rooted in a discovery request lodged under the lawsuit in which The New York Times and other media organizations contend that ChatGPT – and by extension OpenAI’s underlying large language models – inappropriately used their copyrighted content in training, producing outputs that replicate or summarize pay-walled reporting. The plaintiffs claim that access to user chat logs will reveal whether and how often ChatGPT re-generates full or partial passages from their copyrighted articles.
Judge Wang rejected OpenAI’s privacy objections, noting that the company is required to strip out names, email addresses, phone numbers, and any other identifying details before handing over the logs. She added that, in addition to this de-identification, the court has already implemented “multiple layers of protection” to safeguard user information.
OpenAI has been ordered to publish the content seven days after anonymization is complete. The company had earlier warned that complying could set a troubling precedent, noting that the overwhelming majority of the conversations have nothing to do with the lawsuit.
Media Houses Warn Logs Will Uncover Misuse of AI
The New York Times and newspapers owned by MediaNews Group contend that the logs will establish that OpenAI harvested and used their journalism without permission.
They claim that ChatGPT has generated several paragraphs that closely resemble or echo their narratives nearly verbatim. Frank Pine, executive editor of MediaNews Group, slammed the AI company. He said OpenAI was hallucinating when they thought they could get away with withholding evidence about how their business model relies on stealing from hardworking journalists. The outlets also denied OpenAI’s assertion that it had manipulated the AI to infringe on content.
The logs, they also say, will serve as evidence that the chatbot could already create copyrighted content before and without any user input. The case is not about curbing AI innovation, The New York Times said, but rather about ensuring payment and fairness to journalism. OpenAI has appealed Judge Wang’s ruling to U.S. District Judge Sidney Stein, who is overseeing the case.
The company argues that user privacy should not be compromised in court, and any disclosure of these logs, even anonymized data, might erode trust in the company. Dane Stuckey, the company’s chief information security officer, had previously stated that such demands from multiple media organizations contradict common-sense security practices and disregard longstanding privacy protections.
Critics View OpenAI’s Resistance as a Sign of Guilt
Critics argue that OpenAI’s resistance suggests it has something to hide. When AI systems rely on publicly available content, such as journalism, to learn and improve their content, transparency is necessary, they say.
The copyright lawsuit is one among several being brought against technology heavyweights like Microsoft and Meta. Taken together, the cases raise significant questions about how organizations using AI collect, share, and monetize information, as well as to whom, to what extent, and when to ask for compensation.
This places OpenAI on a tightrope, balancing user privacy with accusations of massive copyright theft. It also raises larger questions about how these AI user interfaces align with existing laws, intellectual property protections, and social trust.
If the logs indicate that ChatGPT was copying copyrighted news content without permission, that would have huge implications. Not only for OpenAI — but for the future of AI training, media sustainability, and digital rights.
For now, the decision signals that courts are willing to balance AI innovation against legacy protections for intellectual property — even if doing so complicates notions of privacy in the digital age.