In the case In re OpenAI, Inc. Copyright Infringement Litig., No. 25-md-3143 (SHS) (OTW) (S.D.N.Y. Feb. 6, 2026), regarding Magistrate Judge Ona T. Wang’s November 24, 2025 discovery order determining that OpenAI waived the attorney-client privilege with respect to communications in 2022, New York District Judge Sidney H. Stein stated that he “concludes pursuant to Federal Rule of Civil Procedure 72(a) that the November 24 Order is ‘clearly erroneous’ or ‘contrary to law.’”

Case Discussion and Judge’s Ruling

At the center of the dispute were two datasets—known as Books1 and Books2—that OpenAI created from copies of books downloaded from the website Library Genesis (“LibGen”). The datasets were used to train GPT-3 and GPT-3.5 and were later deleted in 2022. During discovery in the copyright litigation, plaintiffs sought information regarding the reasons for that deletion. OpenAI asserted that communications regarding the decision involved legal counsel and were therefore protected by attorney-client privilege.

Magistrate Judge Wang previously ruled that OpenAI had waived privilege over certain communications related to the deletion of the datasets. The ruling relied on three theories: (1) OpenAI had disclosed a non-privileged explanation (“non-use”) for deleting the datasets and therefore waived privilege; (2) OpenAI had made a “moving target” of its privilege assertions during discovery; and (3) OpenAI placed its state of mind at issue by denying that it willfully infringed copyrights.

Judge Stein rejected each of these rationales.

First, Judge Stein held that OpenAI’s prior statements that the datasets were deleted “due to non-use” did not reveal privileged communications and therefore could not constitute a waiver of attorney-client privilege. Judge Stein explained that attorney-client privilege protects confidential communications made for the purpose of obtaining or providing legal advice. However, merely stating a factual explanation—such as the datasets being deleted because they were not used—does not reveal legal advice. As Judge Stein explained, “OpenAI’s statements that Books1 and Books2 were deleted ‘due to non-use’ do not reveal any legal advice from an attorney to OpenAI.” Because no privileged communication was disclosed, Judge Stein concluded that “those statements cannot be a basis for finding that OpenAI waived the attorney-client privilege.”

Second, Judge Stein rejected the argument that OpenAI had forfeited privilege by shifting its litigation position. Magistrate Judge Wang had concluded that OpenAI’s privilege assertions evolved during the discovery process in a way that justified waiver. Judge Stein disagreed, finding that OpenAI consistently maintained that communications with its attorneys about the deletion decision were privileged. While Judge Stein acknowledged that some of OpenAI’s statements during depositions and briefing may have been “inartful,” he determined that they did not amount to the type of strategic inconsistency that courts treat as a waiver sanction. In other words, the record did not show that OpenAI manipulated privilege claims to gain a litigation advantage.

Third, Judge Stein addressed whether OpenAI had placed its state of mind “at issue” by denying plaintiffs’ allegations that it willfully infringed copyrights. Under the law of the Second Circuit, a party may waive attorney-client privilege if it relies on privileged communications to support a claim or defense—such as asserting a good-faith belief that its conduct was lawful. Magistrate Judge Wang had concluded that OpenAI effectively put its good faith at issue simply by denying willful infringement.

Judge Stein rejected that reasoning. Judge Stein emphasized that denying allegations of willfulness is not equivalent to asserting a good-faith defense. As he explained, “there is a distinction between a copyright defendant merely denying allegations of willfulness—on which a plaintiff bears the burden of proof—and a copyright defendant affirmatively asserting its good faith belief that its actions were lawful.”

Judge Stein further noted that OpenAI expressly disclaimed any intent to rely on a good-faith defense based on legal advice. Instead, OpenAI indicated that it intended to defend the case on other grounds, such as fair use and lack of liability. Because OpenAI was not relying on privileged communications to support a defense, there was no basis for finding an “at-issue” waiver.

In reaching this conclusion, Judge Stein relied on established Second Circuit precedent, which holds that waiver occurs only when a party relies on attorney advice to prove a claim or defense. As he explained, “to effect an at-issue waiver, a party must rely on privileged advice from his counsel to make his claim or defense.” Since OpenAI explicitly declined to do so, the privilege remained intact.

Judge Stein also addressed plaintiffs’ argument that the crime-fraud exception should override the privilege. The crime-fraud doctrine allows courts to compel disclosure of otherwise privileged communications if they were made in furtherance of a crime or fraud. Magistrate Judge Wang had previously rejected the application of the exception after reviewing sample documents in camera.

Judge Stein agreed with that determination. He found that plaintiffs failed to demonstrate probable cause that the communications about deleting the datasets were intended to conceal wrongdoing or further a criminal act. Judge Stein noted that the communications at issue occurred after the alleged acquisition of pirated books and therefore could not have been made “during and in furtherance of” the alleged infringement. He also rejected the argument that deleting the datasets constituted spoliation sufficient to invoke the exception.

Finally, Judge Stein observed that OpenAI ultimately produced versions of the Books1 and Books2 datasets that it was able to recover during the litigation, further undermining claims that the deletion was part of a scheme to conceal evidence.

After analyzing each of the magistrate judge’s rationales, Judge Stein concluded that they were legally incorrect under the governing standard of review. Under Federal Rule of Civil Procedure 72(a), a district court may set aside a magistrate judge’s non-dispositive order if it is “clearly erroneous or contrary to law.” Applying that standard, Judge Stein determined that each basis for the waiver finding failed under applicable privilege doctrine.

Accordingly, Judge Stein vacated the prior discovery ruling. As Judge Stein stated: “Each of the bases for the November 24 Order’s finding that OpenAI waived the attorney-client privilege … is clearly erroneous or contrary to law.” Judge Stein therefore held that the attorney-client privilege remained intact, setting aside the November 24 Order.

So, what do you think? Do you agree that the “Order is ‘clearly erroneous’ or ‘contrary to law.’“? Please share any comments you might have or if you’d like to know more about a particular topic.

Case opinion link courtesy of Minerva26, an Affinity partner of eDiscovery Today.

Disclaimer: The views represented herein are exclusively the views of the author, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.