Artists win first major copyright case against AI companies; Anthropic pays $1.5 billion（約2400億円）

In January 2023, illustrator and cartoonist Sarah Andersen, along with visual artists Karla Ortiz and Kelly McKernan and others, filed a class action suit against Stability, Midjourney, DeviantArt, and Runway AI. Andersen, known for her webcomic Sarah's Scribbles, described her work as deeply personal—"a complex culmination of my education, the comics I devoured as a child, and the many small choices that make up the sum of my life." These and other artists discovered their work had been used to train generative AI systems without consent. The lawsuit was filed just months after Stability's Stable Diffusion and Midjourney were first released, when generative AI was still largely viewed as a curiosity. By 2024, the landscape had changed dramatically, and a wave of legal actions followed.

The most significant resolution so far came in Bartz v. Anthropic, brought by novelist Andrea Bartz and represented by the law firm Susman Godfrey. When The Atlantic published a searchable dataset of training materials, artists including author Kirk Wallace Johnson discovered their published books had been pirated and fed into chatbots. Johnson describes feeling a "cocktail" of emotions: "anger over the brazenness of the theft, worry over what this means for writers, and a healthy thirst for revenge on these massive corporations that have become galactically wealthy" using his intellectual property. Bartz, author of We Were Never Here and The Spare Room, told The Verge: "I felt violated, shocked, alarmed. I had a big emotional response to seeing that something I'd worked on for so many years and poured my heart and soul into was just one of hundreds of thousands or maybe millions of books that these Big Tech companies had just stolen for training their algorithm."

In its ruling, the court found Anthropic liable for copyright infringement related to pirated ebooks downloaded from the internet. The company was ordered to pay $1.5 billion（約2400億円）—the largest settlement in a copyright case—and to destroy its collection of pirated ebooks. However, Judge William Alsup also ruled that Anthropic's use of millions of legally purchased and scanned books, acquired under the name Project Panama, qualified as fair use because the training process was "quintessentially transformative." Bartz disagreed sharply with this part of the ruling: "I strongly disagree with the judge on that part of the ruling … I very much hope that future courts will see the light. Even a library can't buy a physical copy of a book, scan it, and start lending it out as an ebook." Despite this setback on the fair-use question, she emphasized that the settlement itself was historic: "This is the first major win for creatives against an AI company … Hopefully that will guide us toward guardrails that are much needed in the industry."

Meanwhile, musicians face their own battles. Sam Kogon, lead plaintiff in a suit against Google's Lyria AI music engine, accused Google of misusing its own terms of service and improperly leveraging its Content ID system and YouTube data. "They're devaluing our work," Kogon says. "They're giving it away to people, for now, for free. And that's going to disempower a ton of musicians." He characterizes AI music generation as "the most anti-human thing you could do." Google claims its YouTube terms of service grant it broad rights to "reproduce, distribute, [and] prepare derivative works," and the company has filed a motion to dismiss. Entertainment and IP lawyer Krystle Delgado, who represents independent musicians against Suno and Udio, disputes Google's interpretation: "I don't think that anyone uses YouTube thinking that you are giving the rights to remake your content." Yet when she examined YouTube's terms, she found users grant YouTube an "irrevocable perpetual license, meaning you can't ever take it back." Artists view this as an abuse of market power; opting out of YouTube is not a realistic option.

Lawyers and creators remain cautiously optimistic. Delgado believes "the pendulum is swinging in artists' favor." She notes: "Right now with these companies, they're really nervous. Not only were they sued, but the courts and the judges seem to be starting to lean our way, and the court of public opinion too." Polls have shown public support for transparency in AI training. Yet significant risks loom. Johnson emphasizes that the threat is not to famous authors like Agatha Christie but to working creators: "AI could never write The Godfather … But AI could write a mediocre film. AI could write a mediocre book. And there are tons of authors and screenwriters that live in that space." When Meta was sued by novelist Richard Kadrey, comedian Sarah Silverman, and writer Christopher Golden, their filing made this argument explicit: while AI-generated books probably wouldn't displace the work of established authors, they "could very well prevent the next Agatha Christie from getting noticed or selling enough books to keep writing." A federal judge in that case dismissed many initial claims for lack of evidence of market harm, though narrower copyright claims continue to advance. The fundamental concern, shared by all the creators interviewed, is that even legal victories may not protect working artists from the economic disruption AI poses to creative industries.

The Anthropic settlement represents a watershed moment for creators challenging generative AI companies, but the ruling itself contains internal tensions that will shape litigation ahead. Judge William Alsup's decision to hold Anthropic liable for using pirated ebooks affirms that copyright infringement by AI companies is actionable—a claim that had not been tested in court at this scale before. Yet his simultaneous ruling that legally purchased books used in Project Panama qualify as fair use because the training process is "transformative" opens a potential loophole: companies may be able to circumvent copyright concerns by simply purchasing books in bulk and scanning them, rather than stealing them outright. This distinction, which Andrea Bartz explicitly disputes (she notes that "a library can't buy a physical copy of a book, scan it, and start lending it out as an ebook"), suggests that future rulings will hinge on how courts define the scope of "transformative use" in the context of machine learning.

The broader legal landscape remains fractured. While this case sends a clear signal that copyright claims have teeth, other suits have yielded mixed results. The judge in Kadrey v. Meta dismissed many initial claims for lack of evidence of market harm, though narrower copyright-infringement claims proceed. Sarah Andersen's class action against image-generation companies has been "crawling its way through the court system" since January 2023. Meanwhile, Sam Kogon's suit against Google over its Lyria music engine takes a different legal path—challenging whether Google violated its own YouTube terms of service rather than pursuing pure copyright grounds—and Google has filed a motion to dismiss. Lawyers and artists I spoke to believe momentum is shifting in their direction, but the outcome remains uncertain.