Fighting the New York Times’ invasion of user privacy

Twenty million conversations. That's the number the New York Times is asking a court to force OpenAI to hand over as part of its ongoing copyright lawsuit, and OpenAI is not having it. In a blog post, the company frames the request as exactly the kind of privacy invasion its users never signed up for, arguing that people ask ChatGPT everything from medical questions to relationship troubles to half-formed business ideas, and none of that was meant to end up in a discovery folder for a newspaper's lawyers.

The lawsuit itself is old news at this point. The Times sued OpenAI and Microsoft back in late 2023, claiming the companies trained their models on millions of Times articles without permission or payment. What's new is the scope of what the Times now wants as evidence, and OpenAI says that scope has ballooned into something closer to a mass surveillance request than a targeted evidentiary ask.

OpenAI's response has two parts. First, it's fighting the demand directly in court, arguing that sweeping up 20 million private chats is neither necessary nor proportionate to prove or disprove a copyright claim about training data. Second, and maybe more telling, the company says it's using this moment to accelerate security and privacy work that was presumably already on its roadmap, treating the lawsuit as a forcing function rather than just a legal headache.

There's an obvious tension here that OpenAI doesn't shy away from. A company that logs, stores, and in some cases reviews user conversations for safety and training purposes is now positioning itself as the guardian of that same data against outside demands. Whether users find that reassuring or ironic probably depends on how much they already trust OpenAI with their conversations in the first place. Either way, this fight is going to shape how courts think about chatbot logs as evidence for a long time to come, well beyond this one case.