Authors go to war with tech giants over AI

Award-winning author Michael Chabon and fellow writers are suing OpenAI and Meta for allegedly using pirated copies of their works to train their AI chatbots.

Writer Michael Chabon. Picturet: Celeste Sloman for The Wall Street Journal

The award-winning author Michael Chabon and fellow writers are suing OpenAI and Meta for allegedly using pirated copies of their works to train their AI chatbots.

Chabon, who won the Pulitzer prize for his novel The Amazing Adventures of Kavalier & Clay, and the group are suing the tech companies for breach of copyright. They claim that the firms used huge repositories of pirated books to train the AI models ChatGPT (OpenAI) and Llama 2 (Meta).

The other plaintiffs are David Henry Hwang, the Tony and Grammy award-winning playwright, and the authors Ayelet Waldman, Rachel Louise Snyder and Matthew Klam.

ChatGPT and Llama 2 are developed by being trained on billions of words scraped from the internet. Often they used repositories of books and academic papers called “shadow libraries”. While the technology was being developed, often in academic or non-profit settings, authors and artists either were unaware of it or did not object. However, since the release of ChatGPT and the recognition that the technology could be transformational and highly profitable, there has been a backlash from artists.

Michael Chabon with Israeli writer Ayelet Waldman in 2018. Picture: AFP

Similar lawsuits to Chabon’s have been filed against OpenAI and Meta by Sarah Silverman, the comedian, and others. Visual artists have sued OpenAI over the company’s image generator DALL-E.

The Chabon lawsuit against OpenAI claims that it “cast a wide net across the internet to capture the most comprehensive set of content available”, which led it “to capture, download and copy copyrighted written works, plays and articles”. It says that when prompted, ChatGPT accurately summarised The Amazing Adventures of Kavalier & Clay and could write a paragraph in the style of the novel, “imitating Chabon’s writing style”. The lawsuit discusses earlier versions of the chatbot, which cited its data sources, to conclude that “infamous shadow library websites” such as Library Genesis and Z-Library “which host massive collections of pirated books, research papers, and other text-based materials” were used to train ChatGPT.

Last year two Russians who allegedly operated Z-Library were arrested in Argentina at the request of the US justice department. The lawsuit against Meta claims that “shadow libraries have long been of interest to the AI-training community because of the quantity of copyrighted material they contain. For that reason, these shadow libraries are flagrantly illegal”. Both legal actions were filed in California.

Chabon was one of 10,000 authors, including Margaret Atwood, Dan Brown and Jonathan Franzen, who signed an open letter to the chief executives of the main AI companies about the “inherent injustice” of using their words to build “lucrative” AI without consent, credit or compensation.

Maya Shanbhag Lang, president of the Authors Guild, who organised the letter, said at the time: “AI regurgitates what it takes in. It’s only fair that authors be compensated for having ‘fed’ AI. Our work cannot be used without consent, credit and compensation.”

OpenAI has started to address the criticism of its data scraping, allowing websites to opt out of being crawled by its automatic tool.

Despite a flurry of lawsuits, legal experts have questioned whether the authors can prove their works were in the shadow libraries and may be hampered by the fact that the AI does not retain copies of the works it uses.

Conversely, when artists do embrace AI, they are finding it hard to have it recognised by the US Copyright Office, which has recently rejected such works because they are not the product of human authorship. OpenAI and Meta were approached for comment.

The Times