Tremblay v. OpenAI, Inc.

Consolidating actions brought by authors claiming their books were used in training OpenAI’s generative artificial intelligence platform, district court dismisses claims for vicarious copyright infringement, violation of Digital Millennium Copyright Act, negligence and unjust enrichment.

OpenAI develops and sells generative artificial intelligence (AI) software, including ChatGPT, which allows users to enter text prompts to which ChatGPT will respond and “simulate human reasoning,” such as answering questions or summarizing books. ChatGPT is “trained” by inputting large amounts of text from various sources. Among those sources were books authored by Paul Tremblay, Sarah Silverman, Richard Kadrey, Christopher Golden and Michael Chabon.

Tremblay, Silverman, Kadrey and Golden brought two “nearly identical” putative class actions against OpenAI and related entities. OpenAI filed identical motions to dismiss, which the court addressed together, consolidating the two cases, along with a third brought by Chabon.

Plaintiffs alleged six causes of action: direct copyright infringement, vicarious copyright infringement, violation of the Digital Millennium Copyright Act (DMCA), unfair competition, negligence and unjust enrichment. OpenAI sought dismissal of all causes of action except direct copyright infringement. 

In the latest in a string of decisions trimming claims against AI platform developers (see our summaries of decisions in Andersen v. Stability AI Ltd. and Kadrey v. Meta Platforms, Inc.), the court dismissed, with leave to amend, plaintiffs’ claims for vicarious copyright infringement, violation of the DMCA, negligence and unjust enrichment. The court allowed plaintiffs’ unfair competition claim to proceed but noted the possibility that, to the extent it alleges the same violations as plaintiffs’ direct copyright infringement claim, it may be preempted by the Copyright Act. 

Although OpenAI’s motion did not seek dismissal of plaintiffs’ direct copyright infringement claim, the court dismissed plaintiffs’ claim for vicarious infringement on the grounds that plaintiffs failed to sufficiently allege direct infringement. The court explained that plaintiffs failed to allege that any particular ChatGPT output is substantially similar to their copyrighted books. The court found insufficient the broad allegation that “every output of the OpenAI Language Models is an infringing derivative work.”

As did the court in Kadrey v. Meta Platforms, Inc., the court here rejected the argument that Range Road Music, Inc. v. East Coast Foods, Inc. permits plaintiffs to circumvent the requirement to allege substantial similarity. A finding of substantial similarity was not necessary in Range Road Music, Inc. because the infringement at issue was public performance of copyrighted songs. By contrast, plaintiffs in this case did not allege that ChatGPT outputs consist of or contain copies of their books.

Plaintiffs alleged two violations of the DMCA—intentional removal of copyright management information (CMI) under 17 U.S.C. § 1202(b)(1) and distribution of works/copies without CMI under 17 U.S.C. § 1202(b)(3). The court found plaintiffs’ allegation that OpenAI removed CMI was inconsistent with the excerpts of ChatGPT outputs in plaintiffs’ complaints, which contained multiple references to plaintiffs’ names. Further, plaintiffs failed to show how allegedly omitting CMI in copies used to train ChatGPT would give OpenAI “reasonable grounds to know that ChatGPT’s output would induce, enable, facilitate or conceal infringement.” The court rejected plaintiffs’ argument that OpenAI’s failure to reveal which books it used to train ChatGPT constituted evidence of knowingly enabling infringement. As to the distribution of works/copies without CMI, plaintiffs failed to allege that OpenAI distributed their books or copies of their books. Instead, plaintiffs alleged that OpenAI distributed infringing derivatives of their books—without further details as to the nature of these alleged derivatives. 

The court rejected plaintiffs’ unfair competition claim based on allegedly unlawful and fraudulent business practices, as these theories relied on plaintiffs’ DMCA claim, which was dismissed. The court allowed plaintiffs’ unfair competition claim to proceed based on a theory of unfair business practices, assuming that OpenAI used plaintiffs’ books to train ChatGPT for commercial profit. However, in a footnote, the court advised of “the possibility that to the extent the UCL claim alleges the same violations as the copyright claim, it may be preempted by the Copyright Act.”

Plaintiffs’ negligence claim was dismissed on the grounds that plaintiffs failed to allege that defendants owed them a legal duty “to safeguard Plaintiffs’ works” or to “‘maintain[] and control[]’ the public information contained in Plaintiffs’ copyrighted books.” The court also dismissed plaintiffs’ claim for unjust enrichment on the grounds that plaintiffs failed to allege “that OpenAI unjustly obtained benefits from Plaintiffs’ copyrighted works through fraud, mistake, coercion, or request.”

In giving leave to amend, the court noted that a separate order would issue consolidating Tremblay v. OpenAI, Inc., Chabon v. OpenAI, Inc. and Silverman v. OpenAI, Inc., and it directed plaintiffs to consolidate their claims in the amended complaint.

Summary prepared by Melanie Howard and Nathalie Russell