OpenAI is challenging a federal court order that requires the company to preserve all ChatGPT user data indefinitely, including conversations users have deleted. The order stems from an ongoing copyright lawsuit filed by The New York Times against OpenAI and Microsoft.
U.S. Magistrate Judge Ona Wang issued the preservation order on May 13, 2025, after The New York Times requested that OpenAI retain all chat logs that would normally be deleted. The newspaper’s lawyers argue that deleted conversations could contain evidence of users prompting ChatGPT to reproduce copyrighted news articles.
OpenAI complied with the order immediately but did not inform users until June 5, more than three weeks later. The company has now filed an appeal with U.S. District Judge Sidney Stein, arguing that the order conflicts with privacy commitments made to users.
“The New York Times and other plaintiffs have made a sweeping and unnecessary demand in their baseless lawsuit against us: retain consumer ChatGPT and API customer data indefinitely,” said Brad Lightcap, OpenAI’s Chief Operating Officer. “This fundamentally conflicts with the privacy commitments we have made to our users.”
Who is affected by the order
The preservation requirement affects users of ChatGPT Free, Plus, Pro, and Team subscriptions, as well as customers using OpenAI’s API without a Zero Data Retention agreement. When these users delete conversations or use temporary chat mode, their data will now be stored indefinitely instead of being permanently removed within 30 days as previously promised.
ChatGPT Enterprise and ChatGPT Edu customers are not impacted by the order. API customers with Zero Data Retention agreements are also exempt because their data is never stored in the first place.
OpenAI states that retained data is stored separately in a secure system under legal hold. Only a small team of audited legal and security personnel can access the information to meet legal obligations. The data will not be automatically shared with The New York Times or other parties.
Background of the lawsuit
The New York Times sued OpenAI and Microsoft in 2023, alleging that the companies used millions of news articles without permission to train ChatGPT’s underlying language model. The newspaper claims that ChatGPT can reproduce copyrighted content verbatim, potentially harming its business.
In April, Judge Stein ruled that The New York Times had made a sufficient case that OpenAI and Microsoft were responsible for inducing users to infringe copyrights. The judge cited “numerous” and “widely publicized” examples of ChatGPT producing material from Times articles.
Judge Wang granted the preservation order within one day of the Times’ request, agreeing that ChatGPT users might delete conversations to avoid potential legal issues related to the lawsuit.
Privacy concerns and industry implications
The order has raised concerns about user privacy and data retention practices across the AI industry. OpenAI CEO Sam Altman suggested on social media that society needs to develop “AI privilege” similar to attorney-client or doctor-patient confidentiality.
“Talking to an AI should be like talking to a lawyer or a doctor,” Altman wrote on X, formerly Twitter. He called The New York Times’ request “inappropriate” and said it “sets a bad precedent.”
The order may also conflict with European privacy laws, including the General Data Protection Regulation (GDPR), which grants users the “right to be forgotten.” OpenAI acknowledged uncertainty about compliance but stated it must follow the court’s directive.
Some users have expressed frustration on social media after discovering that their deleted conversations are being retained. Security experts warn that the precedent could affect how other AI companies handle user data and privacy commitments.
OpenAI continues to challenge the order and has requested oral arguments. The company hopes that user testimony will persuade the court to set aside the preservation requirement. For now, the timeline for resolving the dispute remains unclear.
Sources: OpenAI, Reuters, VentureBeat, ArsTechnica