Magistrate Judge Wang issued a controversial 28-page opinion concluding that OpenAI waived its attorney-client privilege for communications of in-house attorneys in 2022 related to OpenAI’s deletion of Books 1, 2 datasets. Judge Wang did so on 2 grounds: (1) OpenAI putatively disclosed a “privileged reason” for the deletion based on “non-use” and (2) OpenAI put “at issue” all communications regarding the reasons for deletion by OpenAI’s unwillingness to concede willfulness of its infringement, thereby making OpenAI’s state of mind on willfulness at issue for trial. But Judge Wang rejected (3) the applicability of the crime-fraud exception as well as a similar argument by plaintiffs that OpenAI engaged in misconduct justifying a waiver.
This raises a complex set of legal issues that deserve close scrutiny. I’ve added further thoughts to my original post below explaining why I think Magistrate Judge Wang’s opinion is wrong as a matter of law.
The opinion is internally contradictory. And it makes no attempt to square how (i) OpenAI’s in-house attorney communications in 2022 that were not in furtherance of any crime, fraud, or misconduct because they occurred after the alleged criminal infringement of OpenAI through its researchers torrenting from shadow libraries to assemble Books 1, 2 datasets in 2018 can nonetheless be relevant and used to establish (ii) the willfulness of the very same torrenting activity of AI researchers in 2018. We can also include the OpenAI’s researchers’ use of Books 1, 2 to develop the OpenAI model in 2020.
OpenAI’s in-house attorney communications in 2022 — which occurred several years after the alleged willful infringement in 2018 and 2 years after the alleged willful infringement in 2020 — are not contemporaneous. Because Judge Wang’s opinion (at p. 26) rejected the existence of an ongoing scheme in rejecting the crime-fraud exception, the 2022 attorney communications have no probative value to establish the state of mind or willfulness of OpenAI researchers in 2018 or 2020.
Thus, even if willfulness of the alleged copyright infringement in the 2018 activity and 2020 AI training are “at issue” in the lawsuits against OpenAI, communications by in-house attorneys in 2022 do not bear on the state of mind of the AI researchers who engaged in the activity in 2018 or 2020. Such post-infringement communications by attorneys 2 or more years after the alleged infringement tell us nothing about the state of mind of those OpenAI employees who engaged in the alleged infringement. And that is especially so given that the 3 OpenAI in-house attorneys were not even working at OpenAI before 2021.
In other areas of law involving questions of intent or state of mind, courts have steadfastly rejected the use of evidence not contemporaneous to the defendant’s alleged offense.
For the longer version of my analysis, click the link below: