In a setback for artists, a U.S. District judge dealt a blow to a lawsuit by visual artists on Monday, October 30, 2023. The artists had alleged that Stability AI, Midjourney, and DeviantArt were using their copyrighted work without permission in connection with their generative artificial intelligence training systems.
AI Faces Legal Scrutiny
U.S. District Judge William Orrick has taken action in the case against AI image-generation platforms Midjourney and DeviantArt, which was brought as a proposed class action by Sarah Andersen, Kelly McKernan, and Karla Ortiz. The claims against these AI generators have seen some elements dismissed from the case's canvas. However, there is a glimmer of hope for the artists, as the judge has allowed the artists to refine their case against these two companies.
Orrick found common ground with all three companies in their argument that the images generated by the systems likely did not infringe upon the artists' copyrights. While he is giving the artists a chance to refine their claims, he is not entirely convinced that accusations stemming from the systems' output can stand unless there is a clear case of substantial similarity to the artists' original creations.
Importantly, a direct infringement claim against the third company, Stability AI, has been given the green light for further pursuit. This claim centers around allegations that the company used copyrighted images without proper authorization in the creation of Stable Diffusion. Stability AI vehemently denies the claim that it incorporated these images into its AI system. According to their stance, the process of training their model does not involve wholesale duplication of original works. Instead, it primarily revolves around the development of parameters encompassing elements such as lines, colors, shades, and other attributes related to subjects and concepts found in these works. The resolution of this critical issue, which could significantly impact the case's outcome, remains a subject of dispute.
This is a pivotal issue that will continue to be a topic of extensive discussion, litigation, and potential rulemaking. The crux of the matter is that, under most circumstances, images generated by an AI platform are combinations of interpreted data from their datasets, rather than direct copies of actual images. While this is the claim made by the platforms, I'm not totally convinced. There are instances where a generated image closely matches an actual one. I can only guess this occurs when the dataset lacks adequate image-related bits and descriptions to match a tightly defined prompt. Thus, the issue remains unresolved, partially due to the challenges in obtaining information from the platforms.
Regarding copyright, it is well known that most artists, including fine-art photographers, do not register their images with the U.S. Copyright Office, primarily due to the associated costs. Only registered works have statutory protection in court. While it is true that a work is copyrighted the moment it is created, the absence of registration limits the legal protections available in court. This is not legal advice, so it is advisable to consult a copyright lawyer if you have concerns related to this issue.
You may also want to read my recent post "Essential Tips To Maintain Originality When Using AI Imaging."
Further reading on the case can be found here:
I will continue following the status of AI image generators related to artists' rights and ways to use AI as another tool in your toolbox to create truly original images.