IP Law Daily, COPYRIGHT—N.D. Cal.: Authors Chabon, Hwang bring class action against OpenAi, alleging ChatGPT infringes their copyrighted works, (Sep 12, 2023)

By Robert Margolis, J.D.

Plaintiffs allege that the ChatGPT model trains using unauthorized copies of the authors’ works found in “internet-based book corpora” and creates unauthorized derivative works when prompted by users.

Pulitzer Prize winning author Michael Chabon, Tony Award winning playwright David Henry Hwang and three other authors have sued OpenAI, Inc., and several related entities alleging that the ChatGPT product infringes their copyrighted works by, among other things, including them in datasets used to train the ChatGPT models and by creating derivative works when prompted to write in the style of the authors. The authors purport to sue on behalf of a class of “at least tens of thousands of” authors, and allege that OpenAI has intentionally infringed their copyrights, vicariously done so, and has removed Copyright Management Information in violation of the Digital Millennium Copyright Act. The complaint, filed in federal court in San Francisco, seeks an order for declaratory relief and an injunction temporarily and permanently enjoining OpenAI from continuing its allegedly infringing conduct, including continuing to use and distribute Chat GPT models trained using copyrighted works, and creating and displaying copyrighted works and derivative works. The complaint also seeks statutory and actual damages, restitution of profits, and other available remedies (Chabon v. OpenAI, Inc., filed September 8, 2023).

Alleged infringement. Chabon, Hwang, Matthew Klam, Rachel Louise Snyder, and Ayelet Waldman allege that OpenAI has “incorporated Plaintiffs’ and Class members’ copyrighted works in datasets used to train its [Generative Pretrained Transformer (“GPT”)] models” so that the models will respond to user prompts with “coherent and contextually relevant responses.” They allege that the GPT models are “pre-trained” using collections of copyrighted written works, so that when prompted, ChatGPT “generates not only summaries, but in-depth analyses of the themes present in Plaintiffs’ copyrighted works, which is only possible if the underlying GPT model was trained using Plaintiffs’ works.”

Infringement allegedly occurs first when OpenAI builds the datasets used to train its GPT models by “scraping the internet for text data” and capturing the authors’ copyrighted works. The complaint describes the development and use of datasets that include the “controversial” BookCorpus and two “internet-based book corpora,” Books1 and Books2. The authors allege that on information and belief, the datasets include copyrighted works, which are used without the authors’ permission.

Infringement allegedly also occurs when ChatGPT responds to user queries or prompts, such as to generate writing in the style of a certain author. In response, ChatGPT generates written content based on patterns and connections learned from its analysis of the author’s work within the dataset, which the complaint alleges is a form of a derivative work. Not only is the generated response a derivative work, but the authors allege on information and belief that the model’s ability to create the response means it must have been trained using a dataset that includes an unauthorized copy of the author’s copyrighted work. The complaint includes examples of prompted summaries ChatGPT created for works by each of the named plaintiffs. Plaintiffs allege that OpenAI “benefit[s] commercially and profit[s] handsomely” from the unauthorized use of copyrighted works.

Further, the authors’ works include copyright management information, including the title of the work, the ISBN or copyright registration number, the author’s name, and the year of publication. Chat GPT does not reproduce any of this copyright management information, the authors allege.

Class allegations. The authors purport to bring their claims on behalf of a class of similarly situated authors, which they allege includes “at least tens of thousands of members.” They allege common questions applicable to all of the class members, including (1) whether OpenAI engaged in the conduct alleged; (2) whether the datasets include unauthorized copies of class members’ copyrighted books; (3) whether ChatGPT text responses are infringing derivative works; (4) whether ChatGPT itself is an infringing derivative work; (5) whether OpenAI removed copyright management information; (6) whether injunctive and declaratory relief is proper; and (7) whether damages are proper.

Claims. The authors brought six causes of action: (1) direct copyright infringement, under 17 U.S.C. § 106, et seq., which is alleged to be willful, intentional, and in callous disregard of the authors’ rights ; (2) vicarious copyright infringement, under 17 U.S.C. § 106; (3) removal of copyright management information, in violation of the Digital Millennium Copyright Act, 17 U.S.C. § 1202(B); (4) violations of California’s Unfair Competition Law, Cal. Bus. & Prof. Code §§ 17200, et seq.; (5) negligence; and (6) unjust enrichment.

They seek certification of the case as a class action, declaratory and injunctive relief, statutory and other damages under 17 U.S.C. § 504 for copyright infringement, statutory or actual damages under 17 U.S.C. § 1203, pre-judgment interest, attorney fees, and any other relief the court deems appropriate.

The Case is No. 3:23-cv-04625-PHK.

Attorneys: Daniel Jerome Muller (Ventura Hersey & Muller, LLP) for Michael Chabon, David Henry Hwang and Matthew Klam.

Companies: OpenAI, Inc.; OpenAI, L.P.; OpenAI Opco, LLC; OpenAI GP LLC; OpenAI Startup Fund GP 1, LLC; OpenAI Startup Fund I, LP; OpenAI Startup Fund Management, LLC

News: Copyright TechnologyInternet CaliforniaNews GCNNews