OpenAI pushes back as The New York Times seeks access to millions of ChatGPT logs

OpenAI is gearing up for a crucial battle in its ongoing legal fight with The New York Times (NYT), which wants to dig through a massive trove of ChatGPT logs to look for potential copyright violations. The two sides have jointly asked the court to hold a confidential settlement conference on August 7 — but this isn’t about ending the case. Instead, it’s meant to settle a major dispute over how much access the NYT should get to ChatGPT’s internal records.

According to a report from Ars Technica, this could be OpenAI’s last big attempt to stop the NYT from sifting through what it calls an “extraordinarily broad” range of user data. The company argues that allowing such deep access would threaten user privacy and go against its promise to delete chats after a certain period.

Earlier, OpenAI had accused the NYT of trying to carry out “mass surveillance” of its users. But after losing an earlier round in court, the company shifted tactics and is now trying to limit the number of logs that can be examined. The biggest concern? That private or sensitive conversations users thought were deleted might resurface as part of the legal proceedings.

To find a middle ground, OpenAI proposed a compromise: allow the NYT’s team to analyze a smaller, statistically valid sample of around 20 million ChatGPT logs. This number was backed by computer science researcher Taylor Berg-Kirkpatrick, who said it would be enough to determine whether ChatGPT had reproduced paywalled articles or copyrighted material.