In Re OpenAI, Inc., Copyright Infringement Litigation

Finding that OpenAI waived attorney-client privilege in multidistrict litigation over copyright infringement claims, district court grants plaintiffs' motion to compel OpenAI to produce communications with in-house counsel related to reasons for its deletion of training datasets created from pirated books downloaded from shadow library.

In the consolidated multidistrict litigation (MDL) for copyright infringement against OpenAI Inc. and other defendants, plaintiffs, who are authors and copyright holders of fiction and nonfiction books, alleged that defendants infringed plaintiffs' copyrights by downloading and reproducing plaintiffs' works, using those works to train OpenAI's large language models (LLMs) and creating infringing works in the outputs of OpenAI's consumer-facing AI product, ChatGPT.

Plaintiffs moved to compel OpenAI to produce certain discovery and communications related to the reasons for its deletion of certain training datasets created from pirated copies of plaintiffs' books downloaded from the Library Genesis, or "LibGen," shadow library. Initially, OpenAI claimed that the deletion was due to "non-use." When plaintiffs sought discovery about this reason, however, OpenAI claimed it was covered by attorney-client privilege. In their motion, plaintiffs argued that OpenAI had waived attorney-client privilege by disclosing certain privileged communications (i.e., the reason of non-use) and by putting the reasons for the deletion at issue through its affirmative defense of innocent/non-willful infringement. Plaintiffs also argued that the crime-fraud exception applies to the purportedly privileged communications.

The court granted in part and denied in part plaintiffs' motion to compel, finding that OpenAI had waived the privilege but that the crime-fraud exception did not apply. In particular, the court concluded that, through the representations that OpenAI had made to plaintiffs and on the record, OpenAI had waived privilege over non-use as a reason for its deletion of the datasets. In so finding, the court explained that it "strains credulity" that the reasons for the deletion are privileged because attorney-client privilege does not protect facts. Even if these facts were privileged, however, OpenAI had waived the privilege by making its privilege assertions a "moving target," the court held. Noting that OpenAI first asserted that the datasets were deleted due to non-use and then claimed that all the reasons for deletion were privileged, the court explained that a party cannot identify a reason for its actions and then later assert that the reason is privileged to avoid discovery.

The court also found that OpenAI waived the attorney-client privilege by putting its good faith and state of mind at issue by continuing to dispute plaintiffs' willfulness allegations. OpenAI cannot put its state of mind at issue and simultaneously block discovery into evidence potentially undermining its assertions regarding its state of mind by invoking the privilege, the court reasoned. Although OpenAI stated it would not specifically rely on any particular communications protected by the attorney-client privilege, the court held that this was irrelevant and that OpenAI's use of privilege to block any inquiry into OpenAI's state of mind was improper. As to the scope of the waiver, the court held that because OpenAI waived privilege by putting its good faith and state of mind at issue, OpenAI had waived privilege over all communications with in-house counsel in 2022 related to the reasons for its deletion of (i) the LibGen datasets and (ii) all internal references to LibGen.

The court, however, rejected plaintiffs' argument that the crime-fraud exception applies because plaintiffs had failed to allege a sufficient factual basis to find probable cause to believe that the communications at issue were made during or in furtherance of a fraud or suspected crime, as they were sent only after any potential misconduct occurred. Nor had plaintiffs demonstrated that such communications were intended to conceal such activities.