A judge in California has largely dismissed copyright claims brought by three artists against AI image generators Midjourney, Stable Diffusion, and DeviantArt.
U.S. District Court Judge William H. Orrick failed to find evidence of direct infringements by the AI image companies and mostly granted the defendants’ motion to dismiss the case.
The three artists — Sarah Andersen, Kelly McKernan, and Karla Ortiz — immediately ran into problems as two of them — McKernan and Ortiz — did not register their works with the U.S. Copyright Office.
This left cartoonist and illustrator Andersen — who copyrighted 16 of her works and provided evidence that Stable Diffusion had used her material to train its image generator by showing the results of a search of her name on haveibeentrained.com.
Haveibeentrained.com is a website that allows photographers and artists to see if their work is included in the LAION dataset — a service that provides training data to AI image generators.
Although the case is against three AI image generators, the plaintiffs allege that Midjourney is based on Stable Diffusion and DeviantArt’s “DreamUp” is powered by Stable Diffusion.
The problem for the artists is that the training data for these programs is a black box. Outside of LAION, very little is known about what exactly went into training AI image generators but it is widely assumed that the companies did an almighty scrape of images on the internet which included taking copyrighted and copyrightable pictures.
Judge Orrick writes that it is “unclear” as to whether Stable Diffusion holds “compressed copies” of the images and points to the defense’s argument that the training dataset, which contains five billion images, can “not possibly be compressed into an active program.”
The judge has offered the plaintiffs an opportunity to amend and clarify their theory as to how Stable Diffusion operates its training data.
He wrote that the sheer size of the LAION database may protect the company because 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) copyright training images.”
And since it is almost impossible to produce an identical image that exists within the training data, it will be very difficult for artists to prove that an image that’s come out of Midjourey et al was based on their work.
Not All Bad News For Creators
However, the judge allowed Andersen’s complaint that her 16 copyrighted works were used without her authorization to move forward.
Image credits: Header photo licensed via Depositphotos.