News outlets accuse OpenAI of deleting ChatGPT logs in copyright case

In short: News organizations suing OpenAI say it withheld or destroyed potential evidence, and a federal judge has ordered OpenAI to preserve and share more ChatGPT logs.

What happened

The New York Times, the New York Daily News, and other publishers are suing OpenAI in a consolidated copyright case in federal court in New York. They argue that their articles were used without permission and that ChatGPT can sometimes reproduce or closely paraphrase news content.

A key fight is over ChatGPT “logs,” which are records of user conversations with the chatbot (like a customer service transcript). The publishers say these logs could help show when and how often ChatGPT outputs match copyrighted work, and whether that could hurt their business.

In 2025, the court issued a preservation order telling OpenAI to keep and separate certain ChatGPT output log data that would otherwise be deleted going forward. The Times had argued that OpenAI’s normal deletion practices meant important evidence could be lost. OpenAI said the accusations were wrong and that broad retention creates privacy risks and extra work.

In a separate ruling about what evidence must be turned over, the publishers sought up to 120 million logs. OpenAI proposed a 20 million log sample, but wanted to limit it to conversations it selected as clearly related to the publishers’ works. The judge rejected that approach and ordered OpenAI to produce the full 20 million anonymized logs (with identifying details removed).

Why it matters

This case shows a growing tension between privacy and lawsuits over AI training and outputs. For everyday users, it also highlights that chat logs can become important legal evidence, even when companies prefer to delete them or limit access.

Source: NYTimes