Wang, Ona T., United States Magistrate Judge
Plaintiff,
-against-
OPENAI INC., et al.,
Defendants.
THE NEW YORK TIMES COMPANY.,
Plaintiff,
-against-
MICROSOFT CORPORATION, et al.,
Defendants
ORDER
The Court held a status conference on these matters on December 3, 2024. For the reasons stated on the record, it is hereby ORDERED:
I. Authors Guild, et al. v. OpenAI Inc., et al., No. 23-cv-8292, and consolidated cases (the “Authors Cases”).
A. Plaintiffs’ motion at ECF 270 is DENIED. Discovery and production regarding OpenAI’s large language models is limited to the models identified in response to interrogatory #11.
B. Plaintiffs’ motions at ECF 267 (sealed, unredacted) and ECF 268 (public, redacted) are DENIED. The documents at issue are communications among members of the “Data Working Group,” all of whom are attorneys, and none of whom are document custodians in this case. Defendants represent that nonprivileged documents concerning the same subject matter are likely to be produced from the custodians’ files. Thus it would not be proportional to the needs of the case to direct that Defendants collect, review and log presumptively privileged communications among members of the Data Working Group.
C. The documents sought in ECF 262 appear to relate to public statements from OpenAI employees or former employees concerning “safety” of Gen AI. For the reasons stated in New York Times Co. v. Microsoft Corp., 23-CV-11195 (SHS) (OTW), 2024 WL 4874436 , --- F. Supp.3d --- (S.D.N.Y. Nov. 22, 2024), Plaintiffs’ motion at ECF 262 is DENIED. If, upon meeting and conferring, the parties are able to agree to a limited scope production or proffer, they are directed to so inform the Court in their January 17, 2025, joint chart.
D. Plaintiff’s motions at ECF 193, 230, and 262 are GRANTED in part and DENIED in part as follows: Defendant OpenAI is directed to cross-produce in this case any documents produced in Case No. 23-cv-11195 for Ilya Sutskever, who is a custodian in the Newspaper Cases. If Plaintiffs wish to renew their motion(s) regarding the remaining 7 custodians in dispute, they may only do so after reviewing the Sutskever production and meeting and conferring with opposing counsel.
E. Plaintiffs’ motions at ECF 269 (sealed, unredacted) and ECF 271 (public, redacted) are DENIED. Defendant Microsoft’s production is limited to the documents Microsoft has agreed to produce so far.
II. The New York Times Co. v. Microsoft Corp. et al., 23-cv-11195, and consolidated cases (the “Newspaper Cases”).
A. Plaintiffs’ motions at ECF 305 and 328, which relate to searches of training data sets, are DENIED. The parties are directed to meet and confer on the issues raised in these motions and inform the Court of the status of these issues in their January 17, 2025, joint chart.
B. The remaining “cost shifting” dispute identified at ECF 250 at 4, “VI. Model Inspection Protocol” is resolved as follows: Defendants are directed to provide Plaintiffs one week of “free” searches to assist them with “promot[ing] efficient use of resources.” (ECF 250 at 5). Open AI’s arguments regarding the need for— or their entitlement to—cost-shifting, as made at the conference, were not persuasive because OpenAI has not articulated a cost to them that needs to be shifted. If OpenAI seeks an order shifting costs or limiting, on a dollar basis, subsequent searches by Plaintiffs, they must do so by motion and must articulate actual, calculable costs to them. OpenAI is warned that the Court will consider apportioning costs under Fed. R. Civ. P. 37(a)(5) if OpenAI does not articulate actual, calculable costs.
C. Defendants’ motions at ECF 316 (public, Exhibit J redacted) and ECF 317 (sealed, Exhibit J unredacted)—concerning five additional New York Times custodians—is DENIED. The parties are directed to meet and confer on the issues raised in these motions and inform the Court of the status of these issues in their January 17, 2025, joint chart. As discussed at the conference, the Court will address the issues of additional custodians and documents for each party and each issue separately, assessing the proportionality issues in each dispute on their own. Equality and equity are different concepts.
D. Plaintiffs’ motion at ECF 309 regarding the production of Slack messages and the format of text messages in document production is DENIED as moot.
E. Plaintiffs’ motion at ECF 204 is GRANTED. Defendant OpenAI is directed to add the eight individuals raised in ECF 204 as custodians and the hit count “cap” is raised to 600,000. The parties are further directed to meet and confer whether the 600,000 hit count “cap” applies to all custodial documents produced to date or only those produced after the October 30, 2024, conference. If they elect to raise this issue in the January 17, 2025, joint chart, they should refer the Court to the previously filed documents that inform on this issue.
F. Plaintiffs’ motions at ECF 228 and ECF 313, considered together at the conference, are GRANTED in part. Defendant OpenAI is directed to crossproduce in this case documents produced in the Tremblay case in the Northern District of California for Michael Lampe and Suchir Balaji, who are agreed custodians in Tremblay. ECF 313 is DENIED in all other respects for the reasons stated on the record. The Court reminds the parties that cross-production shall not be counted against the 600,000 hit count cap and encourages the parties to continue working together on finding an acceptable search protocol that yields no more than 600,000 hits.
G. Defendants’ motion at ECF 306 is DENIED as moot. The Court is thankful for the parties’ continuing work to resolve and streamline disputes concerning depositions.
H. Plaintiffs’ motion at ECF 311 is DENIED as moot. The parties are directed to meet and confer on processes to enable Plaintiffs to query the dataset efficiently in advance (and possibly instead) of using a deposition; and inform the Court of the status of these issues in their January 17, 2025, joint chart.
I. Plaintiffs’ motions at ECF 287 and 279 are DENIED. These motions concern the scope of production relating to future, anticipated or later products. While that discovery may be relevant, it is not necessary at this time.
III. Other Issues Pertaining to Both Cases.
A. The interim fact discovery deadline for all cases is hereby EXTENDED to April 30, 2025.
B. The deadline to amend the complaint for all cases is hereby EXTENDED to April 15, 2025.
C. The parties are directed to review the December 3 transcript, which provides additional guidance and colloquy.
D. The parties to each consolidated case are directed not to file any new or supplementary discovery letters or motions between December 21, 2024, and January 10, 2025; any such motion will be stricken and will not be considered. Briefing in support of any mature discovery disputes must be filed between January 10, 2025, and January 16, 2025 and should be done by letter motion and clearly labeled as such. The parties are directed to file a joint chart by January 17, 2025, that identifies (1) all outstanding disputes, (2) each party’s position in no more than two paragraphs, and (3) the relevant ECF numbers (with hyperlinks) for such dispute (including oppositions and replies). If the parties have any updates to the joint chart after January 17, they should be prepared to address those updates orally at the conference.
E. The Court will hold an Omnibus In-Person Status Conference for all cases on Wednesday, January 22, 2025, at 9:30 a.m. in Courtroom 20D, 500 Pearl Street, New York, NY 10007.
The Clerk of Court is respectfully directed to close the following ECF Nos.: (1) Authors Cases (23-cv-8292): 193, 230, 262, 267, 268, 269, 270, and 271; and (2) Newspaper Cases (23- cv-11195): 204, 228, 250, 279, 283, 287, 305, 306, 309, 311, 313, 316, 317, 322, 323, and 328.
SO ORDERED.