The latest fight in the generative AI battleground.
Patent Pending
Google has become the latest target in the fight to protect copyright from AI data scraping, and this time, the group of artists suing the search behemoth has receipts.
Filed in the Northern District of California, the suit brought by photographer Jingna Zhang and illustrators Sarah Andersen, Hope Larson, and Jessica Fink takes Google to task for alleged copyright infringement after using datasets that include their works and those of “billions” of other people.
As the lawsuit details, the plaintiffs discovered that their work was part of Google’s Imagen image generator’s initial training data after Google released a paper about it in 2022 that noted it had used the publicly-available LAION-400M dataset. This “Large-Scale Artificial Intelligence Open Network,” as the name suggests, contains some 400 million images and captions to help train image generators — including, as the artists note in their suit, their copyrighted work.
Just a few months after Imagen was released in beta to a select group of users, Andersen and a pair of other artists sued the makers of the image generators Midjourney and Stable Diffusion for using the LAION-400M dataset without their permission. Like in the Google filing, Andersen and her co-plaintiffs seek to codify image training of their copyrighted works as infringement because they argue that the law hasn’t caught up to the technology, allowing AI companies to classify the scraping of other peoples’ work as “fair use.”
Key Omission
While the Midjourney and Stable Diffusion suit made its way through the courts, getting partially struck down and being re-filed later in 2023, Google released an update to Imagen that did not explicitly state that it used LAION datasets.
As the plaintiffs allege, the company omitted its exact training data source on purpose because it “hoped to avoid being named as a defendant in a lawsuit over the legality of training on mass quantities of copyrighted works without consent, credit, or compensation.” This belief, the suit alleges, is based in part on Google’s hiring of Romain Beaumont, a French AI researcher who was one of the architects of the LAION datasets.
Like other copyright lawsuits against AI companies, including the New York Times‘ suit against OpenAI and Microsoft for allegedly misusing its work, these artists are making a bold philosophical case: that big companies should not be able to use their copyrighted work just because the technology has outpaced the courts.
Because it was just filed, it’ll be a while before these artists or the countless others watching these cases see any resolution or restitution for the alleged misuse of their work — and in the meantime, Google, OpenAI, and every other AI company will keep doing so until the law forces them to stop.
More on AI lawsuits: Drake Pulls Down AI-Generated Diss Track After Lawsuit Threat From Tupac’s Estate
