OpenAI has been ordered to provide millions of anonymized ChatGPT logs to the New York Times, moving a high-profile AI copyright lawsuit into a new and sensitive phase. The ruling puts two issues in direct tension: whether the logs may help evaluate the Times' claims, and whether handing them over creates an unacceptable privacy risk for users who are not parties to the case.
What the court ordered
A federal judge ruled that OpenAI must disclose roughly 20 million anonymized ChatGPT logs to the New York Times. According to Reuters, Judge Ona Wang said on Wednesday that the logs are relevant to two core questions in the dispute: whether OpenAI committed copyright violations, and whether the Times manipulated evidence.
The order does not describe the material as raw public release. The logs are to be anonymized, and the judge pointed to several layers of protections in the case because the data is sensitive. She also said OpenAI's mitigation measures would appropriately address privacy concerns.
OpenAI must provide the cleaned data within seven days. The company immediately appealed the decision.
Why the New York Times wants the chats
The New York Times is seeking access to ChatGPT conversations as part of a lawsuit filed in late 2023. In that case, the paper accuses OpenAI of using its content to train AI models without permission.
OpenAI has said the Times wants to see if users have used the AI to get around its paywall. That makes the logs potentially important to the paper's argument, because they may show how ChatGPT was used in relation to Times content.
The request is also tied to a broader argument over the evidence itself. Judge Wang found that the logs are relevant not only to the copyright accusations, but also to OpenAI's claim that the Times manipulated evidence.
OpenAI's privacy objection
OpenAI is fighting the request on privacy grounds. A spokesperson, citing security chief Dane Stuckey, warned that releasing the logs would disregard longstanding privacy standards and depart from common security practices.
The company's objection is centered on the scope of the demand. OpenAI argued in court that 99.99 percent of the requested transcripts had nothing to do with the accusations. In its view, that means the request reaches far beyond users connected to the lawsuit.
The privacy concern is not new in this dispute. The New York Times had already asked OpenAI to store user content from ChatGPT and its API indefinitely, even if users had deleted their data, in order to preserve evidence.
For OpenAI, the issue is therefore not only whether a specific set of chats may be relevant. It is also whether private ChatGPT conversations from people outside the case should be retained and reviewed because they might contain information useful to the lawsuit.
How the number changed
The scale of the request has shifted during the dispute. The Times originally sought access to 120 million chat logs. OpenAI offered 20 million, saying that anonymizing them would take about 12 weeks.
The court has now ordered OpenAI to provide roughly 20 million anonymized logs. That is a smaller figure than the Times first sought, but it remains a large body of user conversation data.
The numbers matter because they show the practical burden and the privacy stakes at the same time. More logs can mean a broader evidentiary picture, but they also mean more private conversations passing through a process created for litigation.
What the ruling signals
The dispute sits within a broader wave of lawsuits accusing tech companies of using copyrighted works without permission to train their AI models. This case adds another layer: courts may need to examine not only how AI systems were trained, but also how users interacted with those systems afterward.
That creates a complicated balance. The Times argues that ChatGPT logs can help answer questions central to its case. OpenAI argues that the same request reaches users who have nothing to do with the lawsuit and creates a serious invasion of privacy.
Judge Wang's ruling accepts that the data is sensitive, but finds that protections and mitigation measures are enough to let the disclosure proceed. OpenAI's immediate appeal shows that the company is not treating that conclusion as settled.
The result is a legal fight with consequences beyond the number of files involved. It is about what kinds of AI usage data can become evidence when copyright claims reach into the behavior of large language models and their users.