Earlier this year, artist Sarah Andersen, along with other US copyright holders, filed a complaint against Stability AI Ltd, Midjourney and DeviantArt. At the hearing held on July 19, 2023, Judge William Orrick of the US District Court for the Northern District of California had the opportunity to hear oral arguments on defendants’ motions to dismiss and therefore address the legal issues of the case.
Generative AI platforms such as Stability AI (motion to dismiss HERE), Midjourney (HERE) and DeviantArt (HERE) are trained with millions or even billions of images on the Internet and inevitably some of those images are protected by copyright. It’s precisely this kind of use that the Plaintiffs claim to be a violation of their rights as copyright holders. With the development of an ever-increasing number of high quality and well-trained AIs, many cases such as this are arising, but Judge Orrick’s first view could give us a glimpse of the future of AI lawsuits dealing with copyright law.
The Plaintiffs claim, as said, that since the AI companies have trained their models on the artists’ copyrighted works, the output images should be considered as derivative works, and therefore be subjected to the author’s permission. They also allege that the nature of the infringement in this case is more complex than a traditional copyright infringement case. Although they might be right on this last allegation, Judge William Orrick’s reported statements indicated that he agrees with the Defendants’ motion to dismiss the case.
As for the motion to dismiss, the Defendants pointed out various issues with the artists argument and based their motion on multiple allegations. In the first instance, they argue that “a copyright owner must register her works with the Copyright Office before filing a lawsuit for alleged copyright infringement”, whereas Andersen only registered sixteen of her work and McKernan and Ortiz (the other two plaintiffs) have not registered any of their work. Therefore, for them, all the copyright’s claim concerning unregistered works should fail for that reason alone.
They also pointed out the lack of specificity of the Plaintiffs claims: “Plaintiffs’ direct copyright infringement claim based on output images fails for the independent reason that Plaintiffs do not allege a single act of direct infringement, let alone any output that is substantially similar to Plaintiffs’ artwork”, reads the motion. To establish a copyright violation, it is indeed required a comparison between two specific works, been otherwise impossible for the judge to settle the case.
As said, Judge Orrick’s statements suggests he sided more with the AI companies. He agreed with the lack of specificity of the claims and showed skepticism regarding the possibility that an image created by an AI trained on billions of other images could have such an impact on these three artists’ work.
As we are waiting for the final decision, it is likely that anyway the Judge will give the creatives a chance to improve and resubmit their complaint.
Header image created with Microsoft Bing. Prompt: “since the AI companies have trained their models on the artists’ copyrighted works, the output images should be considered as derivative works, and therefore be subjected to the author’s permission”.
