OpenAI is pushing back against a court order that requires the company to hand over 20 million complete user chats to The New York Times and other news organizations. The NYTs had requested the chats, which were allegedly used by users to circumvent the newspaper's paywall, as part of its case against OpenAI for copyright infringement.
However, OpenAI maintains that producing the entire dataset is not necessary and could compromise user privacy. The company claims that only a small subset of these logs would be relevant to the case, while the majority of conversations have nothing to do with the lawsuit. OpenAI argues that revealing complete conversations, which include multiple prompt-output pairs, could expose more private information than individual log entries.
In its filing, OpenAI pointed out that it had previously offered 20 million user chats as part of its defense against copyright infringement claims but was rejected by the NYTs. The company also claimed that producing the full dataset would set a precedent for other companies to demand the production of tens of millions of conversations without first narrowing them down for relevance.
OpenAI is seeking permission from the court to identify which logs are relevant to the case and provide those instead of handing over the entire dataset. The company says it has already implemented security measures, such as client-side encryption, to protect user data.
The NYTs, on the other hand, claims that OpenAI's proposal would not allow them to analyze how real-world users interact with its product and how it delivers news content. The newspaper maintains that the court order is necessary to hold OpenAI accountable for allegedly stealing millions of copyrighted works to create competing products.
A hearing in this case is scheduled for February 26, 2026, which could determine whether the court order stands or if OpenAI gets to produce a smaller subset of relevant logs.
However, OpenAI maintains that producing the entire dataset is not necessary and could compromise user privacy. The company claims that only a small subset of these logs would be relevant to the case, while the majority of conversations have nothing to do with the lawsuit. OpenAI argues that revealing complete conversations, which include multiple prompt-output pairs, could expose more private information than individual log entries.
In its filing, OpenAI pointed out that it had previously offered 20 million user chats as part of its defense against copyright infringement claims but was rejected by the NYTs. The company also claimed that producing the full dataset would set a precedent for other companies to demand the production of tens of millions of conversations without first narrowing them down for relevance.
OpenAI is seeking permission from the court to identify which logs are relevant to the case and provide those instead of handing over the entire dataset. The company says it has already implemented security measures, such as client-side encryption, to protect user data.
The NYTs, on the other hand, claims that OpenAI's proposal would not allow them to analyze how real-world users interact with its product and how it delivers news content. The newspaper maintains that the court order is necessary to hold OpenAI accountable for allegedly stealing millions of copyrighted works to create competing products.
A hearing in this case is scheduled for February 26, 2026, which could determine whether the court order stands or if OpenAI gets to produce a smaller subset of relevant logs.
. I mean, can you imagine having to hand over all those conversations? That's like, so much private info right there 
. I get what the NYTs is trying to do, but come on, it's not worth compromising user privacy for the sake of a lawsuit.
. It's all about setting a precedent and making sure companies don't just start handing over their users' data without a reason 
. And I love that they've already implemented security measures to protect user data – that's some top-notch tech right there 
.
. I'm rooting for OpenAI to win this one and show the court that you don't have to give up all your user info just because someone says so 
.
. Like, they're basically asking for a giant dataset that's basically a treasure trove of private info 
. And OpenAI is right to push back against it - we all know how companies can misuse this kind of data 
to me. What really seems like a stretch is that they're saying that only by seeing all these user chats can they truly understand how their product delivers news content 
.
.
. Like, isnt that gonna be super hard for them to find out what ones are actually important? And whats the point of giving out all that private info anyway? OpenAI is already taking steps to protect users' data, so like, do they really need all that information to prove a point? 
. 20 million conversations are a lot and revealing the entire dataset could be super sensitive info 
. I mean, we're already living in an era where data protection is crucial 
. The NYTs wants access to this info so they can analyze how users interact with their product, but what if some of those interactions are just people chatting about sports or memes 
? Like, does that really contribute to copyright infringement? 
. We should be cautious of that 
. Maybe they can just give the NYTs a smaller subset of logs to look at, like OpenAI is proposing? 
. I mean, can you imagine how much info those 20 million chats contain? Like, it's crazy! 
. It just doesn't seem right to me.