By Ethan Baron, Bay Area News Group
Comedian and author Sarah Silverman is suing social media giant Meta and ChatGPT pioneer OpenAI in a pair of related lawsuits accusing the high-profile California technology companies of breaking state and federal laws by using her memoir “The Bedwetter” as fodder for their artificial intelligence products.
Silverman and two novelists claim Menlo Park-based Meta and San Francisco-based OpenAI violated copyright and other laws by ingesting their books — and those of thousands of other authors — to train their artificial intelligence software.
Both companies copied works protected by copyright law, including books by Silverman, Pennsylvania author Richard Kadrey and Massachusetts writer Christopher Golden “without consent, without credit, and without compensation,” according to the lawsuits filed Friday in U.S. District Court in San Francisco.
OpenAI, which sells ChatGPT via subscription, and Meta, which according to Silverman’s lawsuit is planning to commercialize its LLaMA generative AI models, “not only use those works without permission but then put them into the stream of commerce and the marketplace directly in competition with those who created the works,” said Joseph Saveri, a lawyer for the three authors.
Silverman, Kadrey and Golden and other producers of copyrighted material “run the risk of having their work, their economic livelihood, being entirely supplanted” by AI software,” Saveri said.
Matthew Butterick, another lawyer representing the authors, alleged that OpenAI and Meta have landed on a business strategy of “copyright infringement on a massive, unprecedented scale.”
OpenAI did not respond to requests for comment. Meta declined to comment.
Silverman, Kadrey and Golden are seeking class-action status to bring in all U.S. copyright holders whose work of any kind was used to create the companies’ generative AI products.
“The lawsuits raise some very interesting questions,” said UC Berkeley law school professor Pamela Samuelson. “The courts will take them really seriously, but it’s a little bit early to know how the court’s going to look at it.”
The three authors are among a growing number of creators suing generative AI companies over their use of art, music, photographs and additional copyrighted materials scraped from the internet to create chatbots and other software that produces answers, imagery, music, code, voices and other outputs in response to user prompts. The legal disputes highlight a gap between traditional copyright law and a transformative technology that has swept the world since OpenAI launched ChatGPT late last year.
Courts may decide that using copyrighted material for AI training qualifies as fair use and does not violate copyright law, Samuelson said. Or the U.S. Congress could legislate changes to copyright law to address disputes over AI training data, she said.
The OpenAI lawsuit includes purported prompts and answers related to Silverman’s autobiographical “The Bedwetter: Stories of Courage, Redemption and Pee.” Prompted to summarize the book, ChatGPT produced general information about Silverman’s life — for example that she struggled with bedwetting until her teen years — but also material apparently specific to the book, such as, “Silverman shares details of her relationship with fellow comedian and actor Jimmy Kimmel,” and that she “candidly discusses the impacts of fame on her mental health,” according to the lawsuit.
The court can be expected to focus on whether outputs from the companies’ AI software are “substantially similar” to the creators’ inputs, Samuelson said. “The question is, ‘Would the summary essentially supplant demand for the original?’ ” she said. “Most book reviews have a synopsis from the plot, some details from it, maybe a quote or two.”
Both lawsuits are seeking unspecified damages and a court order upholding the rights of creators whose work is used for training AI.
Lawyers Butterick and Saveri are representing plaintiffs in a number of lawsuits against generative AI companies and say the Silverman cases will not be the last because the firms are gathering vast amounts of copyrighted content from the internet. Both Samuelson and Butterick said licensing of creators’ work could provide a way forward.
“All of human culture has already been scraped up, and we have to kind of decide what we’re going to do next,” Butterick said. “We’re not trying to stand athwart of AI. We just want it to be fair and ethical.”
