Court Orders OpenAI to Permanently Store All ChatGPT Logs, Including Deleted User Conversations

A court has ruled that OpenAI must retain all ChatGPT logs “indefinitely,” including deleted conversations from millions of users. This decision aims to preserve potential evidence in a copyright infringement lawsuit filed by news organizations.

The ruling was challenged by two users of the chatbot, but Judge Ona Wang denied their request.

One of the users, Aidan Hunt, claimed he occasionally uses ChatGPT and shares “extremely confidential personal and commercial information” with the bot. He argued that the data retention order effectively established a “nationwide mass surveillance program,” affecting and potentially harming “all ChatGPT users” who were not informed that their deleted and anonymous chats would still be stored. Hunt warned that retaining AI output data poses the same risks as keeping user input, since the generated information «essentially reveals and often explicitly recounts the queries or topics presented.» He urged the judge to reconsider the order, incorporating exceptions for any chats discussing medical, financial, legal, and personal subjects containing highly sensitive user information that bore no relevance to the interests of the plaintiff news organizations.

According to Hunt, he only learned about ChatGPT’s data retention practices—contrary to the company’s stated policies—by coming across news on an online forum. He speculated that chat data could be obtained by the plaintiffs from news agencies if they found evidence that the deleted chats included attempts by users to craft news articles.

However, the judge responded that there is currently no risk since the chat data had not been disclosed to anyone. Still, Hunt and other users worry that the mere act of data retention could cause serious and irreparable harm. Specifically, Hunt is concerned that OpenAI may not prioritize user privacy if other considerations, such as «financial costs of the case, a desire for swift resolution, and avoidance of reputational damage,» take precedence.

Hunt’s concerns are not unfounded, notes Corinne McSherry, legal director of the digital rights group Electronic Frontier Foundation: “The disclosure order itself poses a real risk to user privacy and sets a precedent in considering numerous other lawsuits across the country. It also represents a broader issue: AI-based chatbots introduce another avenue for corporate surveillance, especially if users lack meaningful control over what happens to their chat histories and records.”

McSherry cautioned that “it is only a matter of time before law enforcement and private plaintiffs start approaching OpenAI to obtain chat histories/user records for various purposes, just as they already do with search histories, social media messages, etc.”

OpenAI will have the opportunity to contest the order on June 26 when Wang hears oral arguments from the creator of ChatGPT.

McSherry believes that legislation needs clarification, as courts are creating confusion in cases involving chatbot users: “All AI chat apps should take steps not only to ensure users can delete their records and be certain they are actually erased but also to provide timely notifications about information disclosure requests.”

In 2023, The New York Times filed a lawsuit against OpenAI and Microsoft, accusing the companies of copyright infringement for training AI on its content. In February 2024, a separate lawsuit with similar allegations was brought by Intercept, Raw Story, and AlterNet.

In response, OpenAI dismissed the lawsuit as baseless. Around the same time, it was reported that the company was negotiating with CNN, Fox, and Time for using materials from these publications to train the ChatGPT chatbot. OpenAI has already signed licensing agreements with Associated Press, Axel Springer, Financial Times, People Dotdash Meredith, and News Corp.

Later, lawyers for The New York Times and Daily News stated that OpenAI engineers accidentally deleted data that could be relevant to the case.