OpenAI Files Notice of Motion to Dismiss 6 Counts in Authors' Copyright Actions

OpenAI will move to dismiss counts II-VI of a copyright infringement complaint, at a hearing Dec. 7 or soon thereafter, said a Monday notice (docket 4:23-cv-03223) in U.S. District Court for Northern California in San Francisco. The notice and memorandum of points and authorities in support of the motion to dismiss address claims brought by authors Paul Tremblay, Sarah Silverman, Christopher Golden and Richard Kadrey.

Plaintiffs allege OpenAI and others “profit richly” from the use of their and class members' copyrighted materials, without having been given consent (see 2306300032). When the ChatGPT chatbot is prompted, it generates summaries of plaintiffs’ copyrighted works, “something only possible if ChatGPT was trained on Plaintiffs’ copyrighted works,” said the complaint. They allege OpenAI infringed their copyrights by removing copyright management information (CMI) from plaintiffs’ infringed works and redistributing those works via ChatGPT without CMI or with false CMI.

The defendant seeks to dismiss counts of copyright infringement, violation of the Digital Millennium Copyright Act and unfair competition under the California Business & Professions Code, negligence and unjust enrichment.

An issue to be decided in OpenAI’s motion to dismiss is whether the claim of vicarious copyright infringement should be dismissed for failure to plead an act of direct infringement or failure to plead facts to support the elements of a vicarious infringement complaint, said the motion, saying a defendant can’t be held vicariously liable unless the plaintiff demonstrated that “direct infringement” actually occurred. The authors’ complaints “fail to identify any bona fide act of direct infringement for which OpenAI could be held liable,” it said.

“Plaintiffs’ theory that 'every output ... is [necessarily] an infringing derivative’ is wrong as a matter of law, which means their vicarious liability claim rests on conduct that is not copyright infringement at all,” said the motion. Claims that OpenAI has the right and ability to control the output of its language models, without supporting facts, lacks detail to sustain a claim, it said, citing Bell Atlantic Corp. v. Twombly.

Plaintiffs’ vicarious liability claim requires that the defendant has a direct financial interest in the infringement at issue, said the motion, citing Perfect 10 v. Giganews. “The operative pleading standard is materially higher, requiring a closer nexus between the infringement of the plaintiffs’ works and a cognizable financial benefit to the defendant,” said the motion. The claim should be dismissed because of lack of facts to support that conclusion, it said.

Plaintiffs allege OpenAI intentionally removed CMI from their infringed works during the AI training process and distributed the models’ outputs without plaintiffs' CMI, in violation of the DMCA, said the motion. The claims are flawed because plaintiffs don’t “plausibly allege that any CMI was removed during the training process,” and they don’t plead facts “sufficient to draw a reasonable inference that OpenAI designed its process with the requisite intent to conceal infringement,” it said.

OpenAI referenced the DMCA’s “double-scienter” requirement, saying “simply removing CMI from a copyrighted work does not yield liability.” A Section 1202 plaintiff must also establish that the removal or alteration was done “intentionally” and the acts performed with “reasonable grounds to know ... that [they] will induce, enable, facilitate, or conceal an infringement [of copyright].” The “the incidental removal of CMI does not raise an inference that the defendant knew its actions would conceal infringement,” it said.

Tremblay's and Silverman’s complaints allege their books were copied by OpenAI “in their entirety” and ingested by OpenAI’s language model, said the motion. The complaints don’t say how OpenAI might delete author names and publication years from the books in its training data, why OpenAI would do that or the basis for the allegation, it said. The “bare speculation that this happened, for no apparent reason and in contravention of the allegation that OpenAI trained its models using entire books obtained from third party sources, … is precisely the kind of ‘unwarranted deduction[] of fact' that this Court must disregard on a motion to dismiss,” it said, citing Gilead Sciences Securities Litigation.

Tremblay’s and Silverman’s unfair competition (UCL) claims are based on the California UCL’s “unlawful” prong, which borrows from other laws and treats them as unlawful practices, said the motion. The claim “fails three times over,” it said. Plaintiffs have not alleged an economic injury caused by the alleged violation, shown an economic injury caused by the violation or pled facts that would justify relief under the UCL, it said.

On the negligence count, plaintiffs’ allegations that infringements were undertaken negligently or “recklessly,” are “irreconcilable with their DMCA claims, which require Plaintiffs to establish that the violations were undertaken ‘intentionally’ or ‘knowing[ly],'" the motion said. None of the allegations suggests OpenAI owed plaintiffs a duty of care, it said.

In their claim of unjust enrichment, plaintiffs provide no explanation “as to what ‘benefit’ they quasi-contractually ‘conferred’ on OpenAI, or how their choice to confer that benefit was the result of ‘mistake, fraud, or coercion,’” said the motion, citing Bittel Technology v. Bittel USA. The complaint only asserts “without explanation, that OpenAI ‘derived profit and other benefits from the use of the Infringed Materials’ and that it would be ‘unjust’" to retain those benefits, it said.