Getty Images filed suit in Delaware federal court, alleging trademark infringement by Stable Diffusion, an artificial intelligence software from Stability AI. Stable Diffusion produces images by processing written descriptions submitted by human users, then recreating images based on its training on images “scraped” from the internet. This type of machine learning must borrow to some extent from existing work; will courts consider this kind of training infringement, and how will it be enforced?
What is Open AI? ChatGPT? And Stable Diffusion?
Artificial intelligence software released in the past two years has many applications in technology, marketing, and other areas of business. Based on human requests, ChatGPT writes content, and Stable Diffusion creates new images. These machine learning technologies rely on large amounts of existing writing and images to “train” them to predict what output should come after a given statement from a human. Frequently, this existing content comes from the internet, where copyrights and trademarks are not always respected. Should their need for this training data negate trademark rights?
Getty and Stable Diffusion
The Getty Images case is one example arising from data scraping used by AI software developers. Getty alleges that Stability AI copied over 12 million images, SEO content, and descriptions to train Stability Diffusion. Additionally, Getty alleged that since Stability AI deliberately copied these images, the images it produced sometimes included “a modified version of a Getty Images watermark, creating confusion as to the source of the images and falsely implying an association with Getty Images.” In addition to more traditional injunctive damages for copyright infringement, Getty Images is asking that the versions of Stability Diffusion that were trained with Getty Images be destroyed.
Can Chat GPT, Stability AI and DALL-E Produce Protectable Content?
This litigation asks a novel legal question: Can AI companies use protected intellectual property as training data? Another legal question many experts are considering is whether content generated by AI can be protected–and if so, who owns those rights? As AI technology increasingly enters the commercial world, this is not yet clear. Are those who purchase licenses to these technologies owners of the resulting intellectual property, or is it the maker of the AI, or even the AI itself?
Chat GPT, Stability AI, DALL-E, and Trademark Litigation and Consumer Surveys
In the type of lawsuit Getty Images filed, holders of protected material would in theory be able to use a consumer survey to measure whether consumers are confused as to the source of the image. Similarly, a dispute over a machine learning technology that was trained on published books to generate new text might also need this kind of survey. This case is in its infancy and we have no answers yet, but where there are questions of consumer confusion, there may be answers in consumer surveys. If you’re interested in discussing how a consumer survey from MMR Strategy Group can help you make your case, contact us today.
To read more blogs regarding Trademark Infringement click here.