OpenAI has appealed a court order from the Southern District of New York requiring it to retain all consumer data for potential use in a copyright law case.
Conversations with ChatGPT using an Enterprise or Edu subscription are not subject to the lawsuit. ChatGPT users who select the Zero Data Retention agreement will also not have their data exposed, regardless of the outcome of the case.
Deleted chat and API content, which would typically be deleted after 30 days, will also be preserved.
Court order sequesters most ChatGPT output into a secure system
The ruling comes after the latest development in the copyright claim that The New York Times brought against OpenAI and Microsoft in 2023.
The newspaper alleged that a significant portion of the AI training corpus consisted of its articles. The case highlighted how large language models benefit from news outlets’ content. OpenAI attempted to have certain aspects of the case dismissed, but it was ultimately allowed to proceed in March.
In May, a court agreed with the Times’ request for OpenAI to keep all output log data. Previously, the Times had cited cases of outputs containing near-exact matches for articles.
OpenAI CEO Sam Altman wrote on X on June 5 that the company believes this “was an inappropriate request that sets a bad precedent,” adding that it will appeal the decision and “fight any demand that compromises our users’ privacy; this is a core principle.”
Keeping chats that may contain evidence of The New York Times articles being directly copied risks exposing numerous customer chats that are irrelevant to the copyright case.
OpenAI responded to concerned customers on June 5, saying the preserved data is only accessed by “a small, audited OpenAI legal and security team” in a secure system and that “If plaintiffs continue to push for access in any way, we will fight to protect your privacy at every step.”
OpenAI’s Chief Operating Officer Brad Lightcap called the demand “sweeping and unnecessary… baseless lawsuit against us” in a recent statement. “This fundamentally conflicts with the privacy commitments we have made to our users,” he said. “It abandons long-standing privacy norms and weakens privacy protections.
“We strongly believe this is an overreach by the New York Times. We’re continuing to appeal this order so we can keep putting your trust and privacy first,” he added.
OpenAI also suggested that even its ability to comply with the GDPR might be compromised due to the court order requiring it to keep the data.
Generative AI throws a wrench into copyright law
Meanwhile, preserving the chats could potentially help The New York Times’ case if their content is found to have been copied too closely.
Thomson Reuters defended its copyright against Ross Intelligence, a legal research firm, in a case involving non-generative AI scraping earlier this year. In March, over 400 actors, directors, writers, and other creators urged US policymakers to address AI being trained on copyrighted work.
Read eWeek’s coverage of the AI licensing deal between The New York Times and Amazon.