Judge Makes Crucial Ruling in AI-Generated Image Class Action Lawsuit

BY Andrew Rossow

October 31, 2023

Last week’s court ruling in favor of Yuga Labs against street artist Ryder Ripps, who must now pay $1.5 million in damages to Yuga for his infringing RR/BAYC NFT collection, was a huge step forward in better distinguishing the relationship between an original work that may or may not be copyrighted or trademarked, and a derivative work that claims to be ‘fair use’ under the First Amendment to the U.S. Constitution.

In the above case, the court validated Yuga’s claim of ownership in its Bored Ape Yacht Club trademarks (BAYC Marks). 

While the BAYC Marks were unregistered with the U.S. Copyright Office (USCO), the court concluded that they were still protected because of how frequently they were utilized in the stream of commerce – including online gaming, IRL events, merchandise marketing, product introductions, and partnerships/collaborations.

But, what happens when an AI image generator is involved and accused of copyright infringement?  

Enter the ongoing class-action lawsuit brought by three artists – Sarah Andersen, Kelly McKernan, and Karla Ortiz – against Stability AI, Midjourney, and DeviantArt. 

Obviously, the integration of AI into the realm of digital art creation has sparked significant debate, particularly when issues of alleged copyright infringement arise. 

In the ongoing case of Sarah Andersen, et al. v Stability AI LTD., et al., U.S. District Court Judge William H. Orrick made a pivotal decision that could very well likely set future precedents in cases involving AI-generated art. 

What’s the Issue?

In the initial complaint, filed earlier this year on January 13, 2023, the three artists alleged that their works had been used without consent.  

They claimed some of their images were included in the Large-scale Artificial Intelligence Open Network (LAION) database, which the accused AI art generators used for training. This database, created by computer scientist and machine learning researcher Christoph Schuhmann and his team, contains billions of images and forms the backbone of these AI systems.

The court’s ruling, filed on Oct. 30 in response to the Defendants’ respective April 18, 2023 motions to dismiss, not only stated that the initial complaint filed by the three artists was defective, but it was “defective in numerous respects,” thereby granting the majority of claims in the Defendants’ motions to dismiss.

A critical factor in Judge Orrick’s judgment was that two of the plaintiffs, McKernan and Ortiz, hadn’t registered copyrights for their art with the U.S. Copyright Office.

Andersen, on the other hand, had copyrighted only 16 of her hundreds of works cited in the complaint but didn’t specifically identify which of those copyrighted works were used as Training Images.

Collectively, the defendants alleged that Andersen, specifically, should be “required to identify which specific works from her registered collections she believes were copied into the LAION datasets” that were eventually used as Training Images for Stable Diffusion.

And Judge Orrick agreed:

“The other problem for plaintiffs is that it is simply not plausible that every Training Image used to train Stable Diffusion was copyrighted (as opposed to copyrightable), or that all DeviantArt users’ Output Images rely upon (theoretically) copyrighted Training Images, and therefore all Output images are derivative images.

Even if that clarity is provided and even if plaintiffs narrow their allegations to limit them to Output Images that draw upon Training Images based upon copyrighted images, I am not convinced that copyright claims based [on a] derivative theory can survive absent ‘substantial similarity’ type allegations. The cases plaintiffs rely on appear to recognize that the alleged infringer’s derivative work must still bear some similarity to the original work or contain the protected elements of the original work,” Judge Orrico ruled in part. 

In other words, the requirement that derivative works must be shown to have “substantial similarity” between the AI-generated image and the original copyrighted work, becomes very challenging to prove unless that AI-generated work clearly references already copyrighted material. 

The Plaintiffs Have 30 Days to Amend Complaint and Cure Defects

Despite this setback for the plaintiffs, Judge Orrick did leave a window open, inviting the artists to amend their claims, and, instead, focus on specific instances of copyrighted image infringement, within 30 days. This also includes curing or fixing any of the deficiencies Judge Orrick outlined in his ruling.

While the majority of claims in the defendants’ respective filed motions to dismiss were granted by Judge Orrick, he did exclude Andersen’s claim alleging direct copyright infringement of her 16 copyrighted works against Stability AI to move forward. 

If one thing’s for certain, the implications of Judge Orrick’s decision resonate beyond this case.

You can stay updated on this case by following the docket here

Editor’s note: This article was written by an nft now staff member in collaboration with OpenAI’s GPT-4.

Dive Deep