IN RE: OPENAI, INC., COPYRIGHT INFRINGEMENT LITIGATION This Document Relates To: 23-cv-11195 25-md-3143 (SHS) (OTW) United States District Court, Southern District of New York Filed May 16, 2025 Wang, Ona T., United States Magistrate Judge ORDER The Court is in receipt of OpenAI’s motion for reconsideration, (ECF Nos. 39-1, 40), of my May 13, 2025, order directing Defendant 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 (the “May 13 Preservation Order”). (See ECF 33). For the reasons below, OpenAI’s motion is DENIED without prejudice to renewal. I. OPENAI’S MOTION FOR RECONSIDERATION IS DENIED The Court assumes familiarity with the facts of this case. As described in the May 13 Preservation Order, the parties in Case No. 23-cv-11195 raised with the Court and discussed in January 2025 a discovery dispute concerning OpenAI’s deletion of ChatGPT output log data. (See 23-cv-11195, ECF 379); (23-cv-8292, ECF 327, “Tr. Jan. 22, 2025”).[1] The News Plaintiffs reraised this dispute on May 7, 2025. (See ECF Nos. 22-1, 24, 27). In essence, the News Plaintiffs allege that OpenAI has been deleting a significant portion of ChatGPT Free, Pro, and Plus output log data since December 27, 2023. (See ECF 29). In the May 13 Preservation Order, I directed OpenAI to preserve and segregate the output log data due to be deleted, and scheduled an in-person status conference (the “May 27 Spoliation Conference”) regarding this dispute for May 27, 2025. (See ECF 32, 33). On May 15, Defendant OpenAI filed the extant motion for reconsideration of my May 13 Preservation Order. (See ECF Nos. 39-1, 40). OpenAI’s motion for reconsideration is DENIED without prejudice to renewal because OpenAI fails to put forth any facts or law that the Court did not consider and that would compel a different conclusion. See In re Iraq Telecom Limited, 18-MC-458 (LGS) (OTW), 2019 WL 5080007, at *1 (S.D.N.Y. Oct. 10, 2019) (“The movant carries the heavy burden of ‘pointing to controlling decisions or data that the court overlooked—matters, in other words, that might reasonably be expected to alter the conclusion reached by the Court.’”) (quoting Shrader v. CSX Transp. Inc., 70 F.3d 255, 257 (2d Cir. 1995)); Vincent v. Money Store, 03-CV-2876 (JGK), 2011 WL 5977812, at *1 (S.D.N.Y. Nov. 29, 2011) (holding whether to grant a motion for reconsideration is left to the “sound discretion of the district court.”).[2] II. MOTION FOR MODIFICATION In the alternative, the Court construes OpenAI’s filing as a motion for modification of the May 13 Preservation Order, and will address OpenAI’s arguments below. OpenAI advances two principal interrelated arguments in support of its motion for reconsideration. First, OpenAI argues, in essence, that the deleted output log data are not relevant.[3] Second, OpenAI argues that preserving all output log data is not proportional to the needs of this case.[4] As an initial matter, the Court entered the May 13 Preservation Order, in part, because the Court’s question to OpenAI at the January 22 conference—whether they could identify the “legal issue here about why you can’t, as opposed to why you would not” retain and segregate the output log data for the users that asked for their chat logs to be delete—remained unanswered. OpenAI’s recent filing answers this question, representing they are now “taking the steps it can to comply with the Order despite the many practical and engineering challenges compliance entails.” (ECF 40). Having resolved this question, the Court will address OpenAI’s concerns regarding relevance and proportionality. A. Relevance The relevance of ChatGPT output log data as a whole is not in dispute.[5] However, implicit in OpenAI’s argument is that there is no material difference between the pool of output log data that is being deleted and the pool of output log data that is being retained; thus, the deletion of output log data marked for deletion in the ordinary course of business would have no effect on the representative sample drawn, and thus no impact on this litigation. The Court explicitly addressed this underlying assumption at the January 22 conference: But hypothetically, say, a ChatGPT user who had been using, found some way to get around the pay wall, right, and was getting The New York Times content somehow as the output, found a way to do it. And then hears about this case and says, Oh, woah, you know, I’m going to ask them to delete all of my searches and not retain any of my searches going forward. I mean, isn’t that directly the problem? (Tr. Jan. 22, 2025, at 38). OpenAI’s letter does not address this open question, while Plaintiffs raise questions about the deletion process that challenge underlying assumptions about the deleted output log data. B. Proportionality OpenAI argues that preserving and segregating the output log data marked for deletion is not proportional to the needs of the case, but this question also remains open. OpenAI has partially addressed this concern by citing technological difficulties and privacy considerations with preservation, and the Court ultimately directed the parties to continue to meet and confer on this issue. (See Tr. Jan. 22, 2025, at 44-46); (see also 23-CV-11195, ECF 441). Accordingly, at present, there is insufficient cause to warrant modification of the May 13 Preservation Order, and the Court expects to discuss with the parties the open questions of relevance and proportionality of preserving and segregating the output log data marked for deletion at the May 27 Spoliation Conference. III. CONCLUSION In sum, OpenAI’s motion for reconsideration is DENIED without prejudice to renewal, and to the extent OpenAI’s letter is a motion to modify the May 13 Preservation Order, such motion is also DENIED without prejudice to renewal. OpenAI may incorporate their renewed motion, if any, in line with their supplemental brief currently due on May 23, 2025. OpenAI’s brief may be up to 15 pages, not including exhibits, if any. The parties should come to the May 27 Spoliation Conference prepared to discuss (1) the underlying assumptions identified above related to the total pool of output log data and the pool of deleted output log data with respect to effects on sampling and (2) the proportionality of preserving and segregating output log data for purposes of this litigation on a going forward basis. The Court appreciates that OpenAI has already begun to take necessary steps to comply with the May 13 Preservation Order to preserve and segregate output log data for purposes of this litigation, notwithstanding the technological difficulties it may entail, and the contractual obligations it may have with its users. The parties are strongly encouraged to continue to meet and confer on this issue, particularly with respect to the technological challenges of preservation and segregation. The parties are also strongly encouraged to include in their briefing proposals to resolve this issue, including potential limitations on going forward preservation or a means to determine whether the retained output logs marked for deletion differ in any material way from the retained logs. SO ORDERED. Footnotes [1] The transcript for the January 22, 2025, discovery status conference that took place before me is currently located at 23-cv-8292, ECF No. 327, and 24-cv-00084, ECF 135. The docket text incorrectly states that the January 22 conference took place before Judge Stein. [2] Reconsideration is an “extraordinary remedy to be employed sparingly in the interests of finality and conservation of scarce judicial resources.” RST (2005) Inc. v. Research in Motion Ltd., 597 F. Supp. 2d 362, 365 (S.D.N.Y. 2009). [3] Specifically, OpenAI argues that the May 13 Preservation Order is “contrary to law” because there is no basis to believe that the data it covers is relevant or that preserving it “will advance this litigation … as opposed to cumulative of the billions of conversations OpenAI has retained.” (ECF 40). [4] OpenAI suggests that because they are preserving “tens of billions of conversations,” the small number of conversations that they have destroyed cannot materially advance Plaintiffs' claims. (ECF 40). OpenAI also argues that it is inappropriate to “disregard [OpenAI’s] users’ privacy interests based on Plaintiffs’ unsupported speculation about what might exist, even though they have not even begun to analyze what does exist.” (Id.). [5] The New York Times’s very first letter motion on this dispute clearly also sets out why the output logs are relevant: that ChatGPT outputs directly infringe the Times’s copyrights. (See Case No. 23-cv-11195, ECF 379) (emphasis added); (see also Tr. Jan. 22, 2025, at 44).