London-based generative AI startup Stability AI has responded to a lawsuit filed in Delaware by stock photo company Getty Images over the alleged misuse of copyrighted images.
In February Getty sued Stability for a “brazen infringement” of its intellectual property, stating that it believed the AI company copied more than 12m copyrighted images — and their associated metadata — without permission.
Getty isn’t the only one suing. Stability is also facing a class action lawsuit in California, filed by artists who argue that text-to-image tools misuse copyrighted works.
Stability’s response
Stability’s opening brief to the court says that the Delaware court lacks jurisdiction over the company’s UK entity, which it argues is a “necessary and indispensable party” to the case.
The response adds that if “even if Getty could establish personal jurisdiction”, the case should be transferred to the northern district of California, where the other lawsuit against Stability has been filed. Stability says the two cases present “substantially overlapping allegations and legal issues”.
The response — which can be read in more detail here — did not directly address the allegations made by Getty.
Fair use
Stability’s founder and CEO Emad Mostaque believes that the way in which generative AI technology transforms the material it uses protects it from these claims of misuse.
Speaking to Sifted in January, Mostaque responded to the first class action lawsuit (also over the misuse of copyrighted material), saying he believes that generative AI technology is protected by “fair use”.
Fair use of copyrighted material, under US law, protects creators who make use of other people’s material if it is “transformative” — generally understood to mean if it changes the nature of that material in some way.
How Stability AI uses images
Stability AI helped fund the open source text-to-image AI model Stable Diffusion and commercialises the technology through a product called DreamStudio.
Stable Diffusion is “trained” on a 100TB dataset that contains 2bn images, including Getty-copyrighted photos.
Christoph Schuhmann, the founder of LAION, the German AI non-profit that built the dataset, says that it’s a “research dataset” and companies that want to use it commercially are responsible for any possible copyright infringements.
“[The dataset] can be used by companies at their own risk and responsibility. We write everywhere on our papers, blog posts and dataset distributions that you should clean them and make sure to use them responsibly, so I cannot speak for Stability AI,” he tellsSifted. “I think if you train an AI model on some data you should make some decisions about what to include in the training data and what not to include. You have to take the responsibility.”
One lawyer specialising in AI copyright, who asked to remain anonymous, tells Sifted that the lawsuit between Getty and Stability is likely to take 18 months to conclude.
