Three artists have filed a copyright infringement lawsuit against several AI art generators, including Stability.ai and Midjourney. The lawsuit alleges that the artists' copyright was violated when Stability.ai and other art generators trained their software using billions of images, which included copyrighted art created by the artists.
In this video, attorney Jake Watson breaks down the arguments in the lawsuit and compares them to related lawsuits and court decisions. He thinks the AI art companies have a good fair use argument for using copyrighted images in its training data.
I think, overall, Stability does have a very good argument for fair use because it appears that their use of the training images and captions in the LAION-5B data set is highly transformative, such that it creates a new purpose for the original images.
I don't think there's much denying that in addition it doesn't appear that this new use by Stability deprives the copyright owners of their right to control and benefit from their original work. However, given that this type of technology has never existed before, especially given how quickly it's developed, I think that as with the dawn of the internet we need to be very careful in considering the far-reaching implications that a finding of fair use might have.
Now, I do think there is a world where people are less inclined to desire the art of the artists who have work in AI image generating data sets. I think that that is a possibility that can exist. I don't think we're in that space right now, but I do think that is a reality and there needs to be an absolute recognition of value that but for the sheer mass of data and the quality of images included in that data Stable Diffusion would not be nearly as good as it is.