Getty files case in US against Stable Diffusion maker for copyright infringement
Stock photo database Getty Images has sued Stable Diffusion maker Stability AI in the US. According to Getty, unauthorized use of his images was made to develop the Stable Diffusion tool. The company previously started a business in the United Kingdom.
In the indictment, filed in a Delaware court, calls Getty Images a “brazen infringement of Getty Images’ intellectual property on a staggering scale.” The stock photo database company claims that Stability AI copied more than twelve million photos from the Getty Images collection, including associated captions and metadata, for the purpose of training the AI model without permission or compensation. Getty states that the purpose of these actions by Stability AI is to build a competitive business. The stock photo database also states that Stable Diffusion’s output often contains a modified version of a Getty Images watermark, which can cause confusion.
Last month, Getty Images filed a case against Stability AI in London’s High Court of Justice, although this has not yet officially started. In accordance with that procedure, Getty first sent a letter to Stability AI, after which the Stable Diffusion maker must respond within a certain period. Regarding this lawsuit, Getty’s CEO said his company wants clarity on the use of copyrighted material for AI purposes. The AI developers usually say that their use falls under fair use.
In any case, the case in the US will revolve around the interpretation of the fair use doctrine. Some artists have already had one before collective action filed against Stability AI, but according to a copyright attorney Getty’s new American case is better structured than the collective case, because Getty focuses on the phase where copyrighted images are inserted to train the model.
Due to concerns about copyright infringement, Getty previously banned the uploading of AI-generated images. Stability AI previously decided to introduce an option for rights holders to opt out of Stable Diffusion so that their work is no longer used to train the model.