In late spring, children’s writer Rachel Vail made a curious discovery: the imaginary worlds she had carefully conjured for her books were not just hers. ChatGPT had access to it too. When prompted, the generative artificial intelligence model, in under a minute, produced an outline for a purported sequel to one of Vail’s bestselling books, ‘Bad Best Friend,’ titling it ‘Redemption.’ Vail is one of many authors who have had their worlds intruded on by technology.
“This felt shocking, it felt like theft,” said Roxana Robinson, a novelist and former president of the Author’s Guild. “These were my characters, my story, and my world. They don’t exist outside of my mind.”
Countless incidents like this led the Author’s Guild, the country’s largest professional writer’s organization, along with 17 authors, including Vail and Robinson, to file a class action lawsuit against OpenAI, ChatGPT’s parent company, in September. The plaintiffs also include popular writers such as George RR Martin, John Grisham, Jodi Picoult, and David Baldacci, all of whom say their books have been used without their consent or payment in the training of the company’s Large Language Models (LLMs). The complaint demands that tech companies pay a licensing fee to authors — $150,000 for each copyrighted book used. For future, the Guild wants a system in place for fair compensation, urging publishers to include an AI related clause.
In its official statement, the Author’s Guild calls OpenAI’s workings identity theft on a large scale. Essentially a copyright infringement suit, the Guild claims that unlicensed eBooks were downloaded from online repositories and included in the training of generative technology. Not only did this rob fiction writers of their work, but it also opened up opportunities for bots to create unauthorized derivative works.
The lawsuit follows a slew of actions taken by writers against artificial intelligence systems this year. In June, Kindle Direct Publishing, where writers can self-publish their books, was criticized for flooding Amazon bestseller lists with AI-generated books. In July, the Author’s Guild wrote an open letter to the CEOs of major companies calling for fair compensation to authors. Signed by more than 15,000 writers, including Margaret Atwood, Dan Brown, Suzanne Collins, and Roxane Gay, it called for fair compensation to the authors. Just last month, the Guild’s screenwriting counterpart, the Writers Guild, succeeded in keeping AI’s fingers out of their work. One of the main demands that was a part of their 148-day-long strike against Hollywood was that producers not use scripts generated or edited by Artificial Intelligence as source material for shows and movies, or force writers to use technology and mandating full disclosure when done so.
“It’s a lot of work to write a book,” said Alexander Greenwood, author of the John Pilate mystery series. “These generative AI folks value these books, or they wouldn’t use them to train. So maybe they can pay a little bit. ChatGPT, if you want to ingest my book, pay a licensing fee. I’d be thrilled.”
At a time when the median full-time author’s salary hovered around $20,000 in 2022, according to the Guild’s latest survey, machines that stand to make billions without paying for the very fabric of the product. This is a primary driving force that made the Guild climb court steps. Going forward, it proposes the creation of a collective organization that licenses, negotiates and distributes royalties to writers who have opted-in to the service.
Vail, also a member of the Guild, noted how OpenAI might insist that what the company has done is not illegal.
“But, we can’t have laws for technologies that don’t exist yet. We may be late to the party already, but we have to hurry up and figure out how to legislate so we build a society that includes creative people,” she said.
Not everyone thinks litigation is the way to solve conflicts in the creative industries, however.
“Let’s say the court settled in favor of authors, maybe you’ve won the fight or you’ve got the moral high ground and you get your pennies,” said Jane Friedman, a writer and publishing consultant. “But I don’t think it changes the trajectory of these companies, of the technology. That ship has sailed.” Earlier in the year, Friedman, too, had made her own curious discovery: five bot-written books were falsely listed as hers on Amazon.
Whether AI is a threat is in the eye of the beholder, Friedman said.
“I think most human beings, when they purchase a book, want it to be by a human author. I don’t see any immediate danger, but obviously there is a lot of anger,” Friedman said.
David Auerbach, a writer and former software engineer, agrees that controlling AI might be hopeless.
“It is like banging on a pinball machine — the world is becoming a pinball machine. We can’t seem to control what AI can do,” he said, clarifying how the trappings of AI systems are opaque.
The authors in the lawsuit entered prompts into the AI models, asking them to produce, or rather reproduce, summaries, final chapters, and unwritten sequels of their works to confirm whether their books had been used ingested as a part of training.
“The authors have this idea that it’s just ripping off their works, smashing everything together. But all traces of the original work are obliterated,” Auerbach said. “They’ve been encoded. You cannot just look into an LLM and see, oh, there’s that book.”
On a panel titled ‘Artificial Intelligence: Will AI Need Us?’ at the Brooklyn Book Festival held in early October, Auerbach claimed that more data now exists than any human can comb through, making the extraction of works that have been ingested from a database almost impossible. “It’s like we are forever closing the barn door after the horse has bolted,” he said.
Robinson remembers feeling disconcerted reading the outline ChatGPT had flashed across her screen. To test if the system had ingested her work, she had prompted it to produce a sequel to her book. “I felt invaded — clearly it had read the book, but had not respected it — it hadn’t bothered to pay attention,” she said.
And now, the authors are paying attention, moving swiftly in their fight against AI. Vail said the Author’s Guild was already on top of copyright infringement by generative models when she approached them in July. The discussions for filing an official suit only began in early September. By the 19th, the Guild had filed its case in the Southern District of New York.
“Part of what I write about for kids is standing up for themselves, even when it’s hard or when the other person has more power. I thought if I’m going to write about that in fiction, I’m going to have to listen to that too,” she said. “When wrong is being done, you have to become a fighter for what’s right.”