A federal judge has allowed a portion of an AI copyright lawsuit against Meta to proceed, rejecting the tech giant’s attempt to have the case dismissed entirely.<\/p>\n
In July 2023, comedienne and The Bedwetter <\/i>author Sarah<\/b> Silverman<\/b>, Sandman Slim<\/i>\u00a0author Richard<\/b> Kadrey<\/b> and\u00a0Ararat<\/i>\u00a0author Christopher Golden<\/b> filed a class-action lawsuit<\/a> against Meta<\/b> and ChatGPT<\/b> maker OpenAI<\/b>, allegedly the tech companies violated copyright laws by training their models on copyrighted books.<\/p>\n Meta, the owner of Facebook<\/strong> and Instagram<\/strong>, launched an AI large language model (LLM) called LLaMA<\/b> in February 2023. Silverman and the two other authors alleged that \u201ctheir copyrighted materials were copied and ingested as part of training LLaMA.\u201d<\/p>\n \u201cMany of Plaintiffs\u2019 copyrighted books appear in the dataset that Meta has admitted to using to train LLaMA,\u201d according to the 2023 court filing, which can be read in full here<\/a>.<\/p>\n In a ruling filed<\/a> on Friday (March 7), US District Judge Vince Chhabria <\/b>granted in part and denied in part Meta’s motion to dismiss the lawsuit. The judge ruled that the authors’ claim under the Digital Millennium Copyright Act<\/b> (DMCA) over alleged removal of copyright management information (CMI) can go forward.<\/p>\n \u201cThe plaintiffs have adequately alleged that Meta intentionally removed [copyright management information] to conceal copyright infringement,\u201d Judge Chhabria wrote.<\/p>\n \u201cThe plaintiffs allege that Meta ‘knew that Llama was especially \u2018prone\u2019 to memorizing and generating outputs of CMI unless CMI was removed from’ its training data. They also allege that Meta took a number of other steps to reduce the likelihood that Llama would generate outputs that would reveal or indicate that copyrighted material was included in training datasets.\u201d<\/p>\n \u201cThe plaintiffs have adequately alleged that Meta intentionally removed CMI to conceal copyright infringement.”<\/p>\n US District Judge Vince Chhabria<\/b><\/span><\/p>\n<\/blockquote>\n The judge added that the authors\u2019 allegations raise a \u201creasonable, if not particularly strong, inference\u201d that Meta\u2019s removal of CMI was an attempt to hide Llama’s potential copyright infringement.<\/p>\n However, Judge Chhabria dismissed the authors’ claim under the California Comprehensive Computer Data Access and Fraud Act<\/b> (CDAFA), finding it’s preempted by federal copyright law.<\/p>\n The judge noted that the authors did not allege that Meta accessed their computers or servers, \u201conly their data (in the form of their books).\u201d<\/p>\n \u201cAnd the only reason that the books Meta acquired could plausibly be considered the plaintiffs\u2019 data is because the plaintiffs own the copyrights on those books,\u201d he wrote.<\/p>\n Further, Judge Chhabria rejected Meta’s argument that the authors lacked standing because the DMCA protects different interests than traditional copyright law. He ruled that “whether the DMCA seeks to prevent the exact injuries the plaintiffs allege is a separate question from whether the plaintiffs have been concretely injured by Meta’s CMI removal.”<\/p>\n Aside from Meta, the authors also sued OpenAI in 2023, with lawyers for Silverman arguing that \u201cmuch of the material in OpenAI\u2019s training datasets\u2026 comes from copyrighted works \u2013 including books written by plaintiffs \u2013 that were copied by OpenAI without consent, without credit, and without compensation.\u201d<\/p>\n The lawsuit against OpenAI argues that, because ChatGPT integrated the copyrighted works into its algorithm, it may itself be a \u201cderivative work\u201d covered by copyright.<\/p>\n \u201cBecause the OpenAI Language Models cannot function without the expressive information extracted from plaintiffs\u2019 works (and others) and retained inside them, the OpenAI Language Models are themselves infringing derivative works, made without plaintiffs\u2019 permission and in violation of their exclusive rights under the Copyright Act,\u201d the complaint states.<\/p>\n In response to the complaint, lawyers for OpenAI argued that the fact that ChatGPT was able to offer accurate summaries of the authors\u2019 copyrighted books doesn\u2019t amount to copyright infringement.<\/p>\n The case represents one of several high-profile legal challenges to AI companies over the use of copyrighted materials in training LLMs. Also in 2023, the Authors Guild<\/b>, which represents 9,000<\/b> professional writers in the US, was joined by 17 of its members, including Game of Thrones <\/i>writer George R.R. Martin<\/b>\u00a0and John Grisham<\/b>, in a lawsuit<\/a> against OpenAI.\u00a0 The suit alleges that OpenAI \u201ccopied plaintiffs\u2019 works wholesale, without permission or consideration,\u201d then fed these texts into its LLMs.<\/p>\n Major music companies have also lodged legal challenges against AI companies. In June 2024, AI music generation platform Suno<\/strong> was sued by the major record companies<\/a>, along with fellow AI music firm Udio,<\/b>\u00a0for allegedly training their systems using the majors\u2019 recordings without permission \u2013 an accusation they\u00a0pretty much admitted<\/a>\u00a0to in court filings in August.<\/p>\n Suno was also sued<\/a> for copyright infringement by GEMA<\/b>, the German collection society and licensing body, in January.<\/p>\n Prior to the Suno and Udio cases, Universal Music Group<\/strong> and other music publishers sued<\/a> AI company Anthropic<\/strong> in 2023, claiming that it trained its AI system\u00a0Claude<\/b>\u00a0on lyrics from at least\u00a0500<\/b>\u00a0songs by major artists,\u00a0including\u00a0Katy Perry,\u00a0<\/b>the\u00a0Rolling Stones,\u00a0<\/b>and\u00a0Beyonc\u00e9<\/b>, and sought\u00a0$150,000<\/b>\u00a0in damages per infringement.<\/p>\n