Sarah Silverman, who’s suing OpenAI and Meta for allegedly infringing upon her writing while training their respective AI offerings. Photo Credit: Gage Skidmore
Last month, Sarah Silverman (who wrote 2010’s The Bedwetter) joined multiple other authors in suing OpenAI for allegedly training ChatGPT on copyrighted writing without authorization. Now, the defendant entity has moved to dismiss the majority of the suit.
OpenAI just recently submitted a motion to dismiss with prejudice five of the plaintiffs’ six causes of action, the excluded claim being for direct copyright infringement, which OpenAI intends “to resolve as a matter of law at a later stage of the case.”
DMN covered the relatively straightforward action in detail soon after its filing – on top of a separate-but-similar complaint against LLaMA developer Meta, which was promptly ordered into “alternative dispute resolution.” In brief, though, Silverman, Ararat writer Christopher Golden, and Sandman Slim author Richard Kadrey accused OpenAI of negligence, violating the Digital Millennium Copyright Act as well as California’s unfair-competition law, and more.
“Those claims, however, misconceive the scope of copyright,” OpenAI wrote of the overarching allegations at hand, “failing to take into account the limitations and exceptions (including fair use) that properly leave room for innovations like the large language models now at the forefront of artificial intelligence.”
Beginning with the suit’s second claim, OpenAI in its dismissal motion expressed the belief that the plaintiffs had failed to describe the direct infringement, right and ability to supervise the alleged infringement, and direct financial interest required to demonstrate vicarious infringement.
“Because the derivative-work theory underlying Plaintiffs’ vicarious liability claim is wrong as a matter of law, the Complaints fail to identify any bona fide act of direct infringement for which OpenAI could be held liable,” the legal text reads, indicating also that the plaintiffs “make no attempt to explain how or why any particular outputs are substantially similar to their books.”
Regarding OpenAI’s alleged violation of the DMCA – and specifically Section 1202(b), concerning the removal of copyright management information like author names and publication dates – the dismissal motion explores a purported failure to demonstrate that the defendant had actually omitted CMI during the ChatGPT training process.
“The Complaints are completely devoid of any explanation as to (1) how OpenAI might delete author names and publication years from the books in its training data, (2) why OpenAI would do such a thing, or (3) what Plaintiffs’ good-faith basis for believing this occurred might consist of,” the motion continues.
Next, shifting to unfair-competition practices under California law, OpenAI made clear its position that the “claim fails three times over” – because of the alleged absence of a proven DMCA violation within the complaint, the alleged lack of “an economic injury flowing directly from the alleged DMCA violations,” and the alleged exclusion of adjacent “facts that would justify any relief.”
Lastly, OpenAI relayed that Sarah Silverman and the other plaintiffs had failed to state claims for unjust enrichment and negligence. “The core facts appear to be nothing more than those underlying the copyright infringement claims,” the text states towards its conclusion, “re-asserted on a quasi-contract theory.”