Federal judge William Orrick dismissed the majority of the copyright infringement claims that 3 artists brought against Stability AI, Midjourney, and DeviantArt.
Earlier this year Sarah Anderson, Kelly McKernan, and Karla Ortiz claimed their copyrights were infringed when their work was used to train the AI models to then generate derivative works.
McKernan and Ortiz had to drop their claims when it emerged that they didn’t register copyrights of their work before suing. This left Anderson as the only plaintiff on the copyright claims.
Midjourney is one of the most popular AI art generators and DeviantArt is an online art community where users can share their AI-generated images. DeviantArt’s DreamUp product also relies on Stable Diffusion to generate images.
The core of the claims focuses on Stability AI’s Stable Diffusion model, with Midjourney and DeviantArt allegedly being complicit for using the model.
That’s not how it works
The judge ruled that the claims were “defective in numerous respects.” A big part of this decision comes down to how the claims were worded and how AI image generators actually work.
The original suit claims that in the process of training Stable Diffusion, “Stability caused those images to be stored at and incorporated into Stable Diffusion as compressed copies.”
The defense argued that this claim wasn’t only false, but that it was impossible for copies of billions of images to be stored in the model. Stability AI argues that training its model involves extracting attributes from images like lines, shades, and colors and developing parameters related to these.
Orrick’s ruling stated that the “Plaintiffs will be required to amend to clarify their theory with respect to compressed copies of Training Images and to state facts in support of how Stable Diffusion — a program that is open source, at least in part — operates with respect to the Training Images.”
We did not lose the case. News outlets saying this are false. Several claims were submitted, the judge dismissed some of them, and we get to amend (update) the ones not dismissed w/ prejudice. We ARE going to trial with Stability over training on copyrighted art. Don’t lose hope.
— Kelly McKernan (@Kelly_McKernan) October 31, 2023
The judge also questioned whether DeviantArt and Midjourney could be liable for direct infringement of copyright if Stable Diffusion “contains only algorithms and instructions that can be applied to the creation of images that include only a few elements of a copyrighted” image.
Orrick also said the plaintiffs needed to be clearer about their claim against Midjourney. He said they needed to clarify whether their claim was “based on Midjourney’s use of Stable Diffusion, or Midjourney’s own independent use of Training Images to train the Midjourney product, or both?”
If the artists want to reintroduce their amended claims they will also need to prove that Midjourney produces substantially similar works to their images.
That’s going to be tricky because Orrick’s ruling said the plaintiffs admit that in general, “none of the Stable Diffusion output images provided in response to a particular Text Prompt is likely to be a close match for any specific image in training data.”
While Midjourney and DeviantArt are off the hook (for now), Stability AI will still need to face Anderson’s direct copyright infringement claim related to the wholesale copying of images to train Stable Diffusion.
Stability AI faces several related lawsuits, including from Getty Images. If the rulings go against them it will have ramifications for other generative models too.
