Artists have been hit hard by their ongoing legal battle with generative AI developers. On Monday, a US district judge ruled that a lawsuit against AI-image generator Midjourney and digital art community platform DeviantArt could not proceed, saying the plaintiffs did not provide sufficient evidence to support claims of copyright infringement.
Judge William Orrick granted the defendant's motion to dismiss the copyright infringement claim as “defective in many respects.”
While the lawsuits against Midjourney and DeviantArt were dismissed, Orrick, however, allowed a separate infringement claim by AI developer Stability AI, creator of Stable Distribution, against illustrator Sarah Anderson to move forward.
Generative AI refers to artificial intelligence programs that prompt the user to generate text, images, videos, and audio. Generative AI models are trained on large amounts of data from various sources, including the Internet.
In the DeviantArt case — an online social network for artists and art enthusiasts — Orrick said the plaintiffs failed to show how the platform was at fault for collecting Internet content. That was done by LAION, a Germany-based non-profit that develops open-source artificial intelligence models and datasets.
“Plaintiffs have failed to allege certain persuasive facts that DeviantArt played any positive role in creating training images by scraping and using Anderson's and other registered works,” the judge wrote. “The complaint instead alleges that the scraping and creation of training images was done by LAION at Stable's direction and that Stable used the training images to train for Stable Diffusion.”
In the suit, plaintiffs Kelly McKernan, Carla Ortiz and Anderson allege that Stability AI violates not only direct copyright infringement laws, but also the Digital Millennium Copyright Act (DMCA), artists' rights of publicity and platforms' terms of service. Orrick admits they may have an issue on that front.
Plaintiffs have sufficiently committed direct infringement by alleging that they “downloaded or took billions of copyrighted images without permission to create Stable Diffusion” and used those images to train Stable Diffusion and cause those images to be stored and embedded. to Stable Diffusion as compressed copies,” says Orrick.
The plaintiffs also said that art created in the same artist's style – perhaps billions of variations – also constitutes copyright infringement.
“Each output image from the system is derived only from hidden images, which are copies of copyrighted images,” the artists said. “For these reasons, every composite image is necessarily a derivative work.”
Stable Diffusion's lawyers instead argued that Anderson should have identified that her copyrighted or “registered” works were copied from training material for Stable Diffusion. The judge explained that this would be difficult to do.
“The complaint is that depending on how the distribution process works, none of the output images ‘will be close to a particular image in the training data,'” he said.
In July, Orrick said the plaintiffs should “better distinguish” the lawsuit and the AI imaging companies. At the time, Orrick said, it was “impossible” for them to handle certain tasks, given the amount of training data.
In today's ruling, he asked the plaintiffs to “improve to clarify their theories” about whether the AI models contain “compressed copies” of their work or how to reconstruct “algorithms and instructions.” But it denied Stablitet's motion to dismiss the direct copyright infringement claim.
Orrick denied Midjourney's motion to dismiss the class action aspect of the case, saying that “it would be premature to deny the parties the opportunity to resolve issues or claims in a class action.”
The lawsuit against AI image creators is one front in the ongoing battle against copyright infringement and AI. In September, several authors, including Game of Thrones creator George RR Martin, joined a lawsuit against OpenAI, the creator of ChatGPT, alleging that their work was recorded on the chatbot's training data. Earlier this month, Universal Music sued cloud AI developer Anthropic for widespread copyright infringement.
“In the process of building and operating the AI models, Anthropopic illegally copied and distributed vast amounts of copyrighted works – including lyrics – and countless musical compositions owned or controlled by publishers,” Universal Music's lawyers said in the lawsuit.
The next hearing in the lawsuit against Stability AI, Midjourney and DeviantArt is set for November 7.
Edited by Ryan Ozawa.
