A federal judge has ordered OpenAI to preserve ChatGPT conversation logs and turn over 20 million of them to news publishers and authors suing the company for copyright infringement, and that ruling is at the center of the ChatGPT chat log preservation lawsuit brought by The New York Times and other plaintiffs. The preservation order overrode OpenAI’s standard 30-day deletion policy for months in 2025, sweeping in conversations from most consumer and API users who had no idea their deleted chats were being held. The logs themselves are now being de-identified and produced to opposing counsel under a January 2026 discovery ruling.
Which ChatGPT Users the Order Covered
Magistrate Judge Ona T. Wang issued the preservation order on May 13, 2025, directing OpenAI to “preserve and segregate all output log data that would otherwise be deleted on a going forward basis until further order of the Court.”1Ars Technica. NYT v. OpenAI Preservation Order It reached ChatGPT Free, Plus, Pro, and Team subscriptions, along with API usage by customers without a Zero Data Retention agreement.2OpenAI. Response to NYT Data Demands
Enterprise and Education customers were excluded, a point the court clarified at a May 27 hearing.2OpenAI. Response to NYT Data Demands Users in the European Economic Area, Switzerland, and the United Kingdom were carved out later, under an October 9, 2025, stipulation.3Ars Technica. Stipulation and Order Terminating Preservation Everyone else who used ChatGPT during the preservation window had their deleted conversations held rather than erased.
Under OpenAI’s normal policies, deleted chats are permanently removed within 30 days.4OpenAI Help Center. Chat and File Retention Policies in ChatGPT The court order froze that process. OpenAI stored the preserved logs in a separate, secured system under legal hold, accessible only to a small, audited internal legal and security team.2OpenAI. Response to NYT Data Demands
Why the Court Ordered Preservation
The order grew out of a copyright suit The New York Times filed against OpenAI and Microsoft on December 27, 2023, in the Southern District of New York. The Times alleged that OpenAI trained its large language models on millions of Times articles and that ChatGPT could reproduce that content verbatim or near-verbatim in user responses.5NYT Company. NYT Complaint Dec 2023 Eight Alden Global Capital newspapers filed a related complaint in April 2024, and the Center for Investigative Reporting followed in June 2024.6Justia. Daily News LP et al. v. Microsoft Corporation et al. On April 3, 2025, the Judicial Panel on Multidistrict Litigation consolidated 12 actions into MDL No. 3143 before Judge Sidney H. Stein, sweeping in class actions by authors including George R.R. Martin, John Grisham, David Baldacci, Jodi Picoult, and the Authors Guild.7GovInfo. MDL Transfer Order, In Re OpenAI Inc. Copyright Infringement Litigation
The plaintiffs argued to Magistrate Judge Wang that users could delete conversations containing evidence of copyright infringement, essentially destroying the record of ChatGPT’s outputs. OpenAI opposed the request, calling it a “carte blanche, preserve everything” demand and pointing to “numerous privacy laws and regulations throughout the country and the world.”1Ars Technica. NYT v. OpenAI Preservation Order Wang initially declined the request in January 2025, then reversed course in May. She denied OpenAI’s motion for reconsideration on May 16, and Judge Stein denied the company’s objections the week of June 23, 2025, pointing to OpenAI’s own terms of service allowing data retention for “legal process.”8Ars Technica. NYT to Start Searching Deleted ChatGPT Logs After Beating OpenAI in Court
OpenAI’s underlying defense in the copyright case is fair use. The company contends that training a language model on copyrighted material is transformative and that ChatGPT’s occasional reproduction of copyrighted text is a “rare bug.”9AI Lawsuit Tracker. New York Times v. OpenAI The company has also accused the Times of manipulating its product, alleging it took “tens of thousands of attempts” to coax ChatGPT into producing the verbatim excerpts attached to the complaint.10Harvard Law School. Does ChatGPT Violate New York Times Copyrights Judge Stein’s April 4, 2025, opinion largely kept the plaintiffs’ direct and contributory infringement claims alive, though it dismissed certain narrower DMCA claims.11U.S. District Court, S.D.N.Y. OpenAI MTD Opinion
The 20 Million Logs Being Produced
Preserving the data was only the first step. In July 2025, news plaintiffs moved to compel production of a sample of 120 million ChatGPT logs. OpenAI countered with a 20-million-log sample, roughly 0.5 percent of its preserved conversations, scrubbed of personally identifiable information.12ABA Journal. ChatGPT Creator Must Turn Over 20M Chat Logs in Copyright Litigation The plaintiffs initially agreed.
OpenAI then tried to narrow the production further, proposing keyword searches that would surface only logs referencing the plaintiffs’ specific works.13Bloomberg Law. OpenAI Must Turn Over 20 Million ChatGPT Logs, Judge Affirms On November 7, 2025, Judge Wang rejected that approach and ordered production of the full 20-million-log sample, covering conversations from December 2022 through November 2024. She reasoned that even logs not mentioning the plaintiffs’ works were relevant to whether ChatGPT’s outputs compete with or substitute for copyrighted content. She denied OpenAI’s reconsideration motion on December 2 and extended the same obligation to the class plaintiffs three days later.14U.S. District Court, S.D.N.Y. NYT v. OpenAI, Order Affirming Discovery
Judge Stein affirmed on January 5, 2026. He acknowledged user privacy concerns were “sincere” but found them mitigated by three safeguards: the sample size reduction from tens of billions of logs to 20 million, mandatory de-identification, and the existing protective order governing discovery. Stein rejected OpenAI’s comparison of the logs to wiretapped calls, writing that ChatGPT users “voluntarily submitted their communications” to the platform.12ABA Journal. ChatGPT Creator Must Turn Over 20M Chat Logs in Copyright Litigation He also declined to impose a “least burdensome discovery possible” standard, finding no case law supporting one.13Bloomberg Law. OpenAI Must Turn Over 20 Million ChatGPT Logs, Judge Affirms
What This Means for Your Privacy
Users typed sensitive queries into ChatGPT expecting deleted conversations to disappear within 30 days. That expectation was overridden for months, and 20 million of those conversations are now moving to opposing counsel.
The court has treated de-identification and the protective order as adequate protection. Researchers have shown that conversational data often contains personally identifiable information, including full names, addresses, and identification numbers, that can survive automated scrubbing.15Jones Walker LLP. Your ChatGPT Chats Are About to Become Evidence The order also permits production of data from millions of users who are not parties to the case, without their notice or consent.
The stakes reach businesses that route customer data through ChatGPT. Legal analysts have warned the retention order may prevent organizations from honoring consumer deletion requests, putting them at odds with data-minimization commitments or international rules like the GDPR.16Bloomberg Law. OpenAI Case Amplifies Legal Tension Between Discovery, Privacy The litigation has also been read as a signal to other AI providers, including Google and Anthropic, to assess their own preservation exposure.
Where the Preserved Data Stands Now
The indefinite preservation obligation ended on September 26, 2025, under a stipulated order allowing OpenAI to resume its standard 30-day deletion practices for new data.2OpenAI. Response to NYT Data Demands The historical logs collected between roughly April and September 2025 remain segregated. The October 9, 2025, stipulation confirmed OpenAI must continue preserving that already-captured data, though it could stop retaining logs from users in the European Economic Area, Switzerland, and the UK.3Ars Technica. Stipulation and Order Terminating Preservation The company must also continue preserving, on a going-forward basis, logs tied to accounts linked to specific domains identified by the news plaintiffs.
OpenAI has said the preserved data “will not be turned over to the New York Times, the Court, or anyone else at this time” beyond the ordered production, and has signaled it may seek appellate review.2OpenAI. Response to NYT Data Demands The 20 million de-identified logs ordered produced in January 2026 are going to both the news and class plaintiffs for expert analysis on market harm and fair use. Briefing on summary judgment concluded April 2, 2026, with a hearing expected around May and a ruling anticipated in the third quarter; a trial, if the claims survive, is projected for 2027.9AI Lawsuit Tracker. New York Times v. OpenAI
The “AI Privilege” Idea
OpenAI CEO Sam Altman responded to the preservation order by proposing a new form of legal protection. On June 5, 2025, he posted on X that “talking to an AI should be like talking to a lawyer or a doctor,” calling for a privilege that would shield user-AI interactions from compelled disclosure.17VentureBeat. Sam Altman Calls for AI Privilege as OpenAI Clarifies Court Order He returned to the idea in a late July 2025 podcast, describing the lack of protection as “very screwed up” given how many users treat ChatGPT as a therapist or life coach.18NYU Shanghai. Sam Altman: ChatGPT Therapy Chats Offer No Legal Confidentiality No legislature or court has recognized such a privilege, and legal commentators have expressed skepticism that one will emerge from the current litigation.