A group of 10 visual artists can pursue copyright claims against four companies that use generative AI to convert text to images, a California court ruled on Monday, August 12.
The artists filed a class action lawsuit against Stability, Midjourney, DeviantArt and Runway in U.S. District Court in Northern California last November, claiming that image generator Stable Diffusion uses their artwork as “training.” The original lawsuit also alleged that the generator, which all four companies use, can mimic the artists’ style, according to court documents. This week’s statement comes after AI companies tried to dismiss the artists’ lawsuits.
“The plaintiffs’ core claims will now be forwarded for discovery and trial,” Matthew Butterick and Joseph Saveri, two of the artists’ attorneys, wrote in a statement to Hyperallergic. The artists who filed the lawsuit are Sarah Andersen, Kelly McKernan, Karla Ortiz, Hawke Southworth, Grzegorz Rutkowski, Gregory Manchess, Gerald Brom, Jingna Zhang, Julia Kaye and Adam Ellis.
While District Judge William Orrick dismissed the plaintiffs’ claims for unjust enrichment and breach of contract, he ruled that the artists’ copyright complaints were plausible, including claims that their artwork was used to train the companies’ AI.
“You can’t copy someone else’s work without their permission and then not keep that copy, because that’s copyright infringement,” said Philippa Loengard, executive director of the Kernochan Center for Law, Media and the Arts at Columbia University. Hyperallergic.
Lawyers for Stability, Midjourney, DeviantArt and Runway did not immediately respond to requests for comment. Hyperallergic to comment on the case. At a hearing in May, Andy Gass, a lawyer for DeviantArt, warned of the “chaos that would be wrought” if the lawsuit were allowed to proceed, adding that DeviantArt “has not developed any new generation AI models… all it has allegedly done is take StabilityAI’s Stable Diffusion model, download it, upload it, and offer a version” to its users.
For artists who say they’ve seen AI image generators mimic their work, such as New York-based illustrator and AI critic Molly Crabapple, models trained on real artwork could completely disrupt their lives as working artists. In May 2023, Crabapple penned an open letter urging the publishing industry to avoid using AI in illustration.
“(These image generators) rely on the stolen work of artists to enrich a few Silicon Valley companies while deskilling, disempowering and ultimately replacing workers,” Crabapple said. Hyperallergic.
The artist shared the result of a request to OpenAI to create an image in her style Hyperallergic. The chatbot, in its artistic style, offered four options for each input she gave.
Earlier this year, an alleged list of artists whose work will be used to train Midjourney’s generator was leaked, including well-known names like David Hockney and Yayoi Kusama. But as image generators learn to imitate the artistic style of illustrators, Crabapple says working-class artists will be most affected.
“These are not people who have made tons of money,” Crabapple said. “These are people in late middle age in America who are approaching retirement and have nothing because these Silicon Valley companies have decided they need to make billions and billions of dollars.”
The class action lawsuit centers on the alleged use of the so-called LAION dataset, a collection of five billion images, by the two companies to develop the Stable Diffusion image generator.
According to Loengard, there is a risk that artists’ works that they distribute online will be “scanned” by AI companies.
“There are all kinds of violations here,” Loengard said. “AI only works if it has data, and AI cannot generate its own data. In order for the platform to do what it does, it needs new material.”
While the artists’ lawyers celebrated the ruling as a “significant” next step in the case, Loengard added that it was too early to say whether this case and other similar copyright cases would protect creative people in the age of artificial intelligence.
