In a pivotal legal battle shaping the AI landscape, the Business and Property Courts of England and Wales have ruled to proceed with a lawsuit filed by Getty Images against Stability AI, the company behind the AI model Stable Diffusion. Stability AI faces allegations of using Getty's copyrighted material without consent to train its AI model, which is accessible in the UK through various means including Dream Studio.
Getty Images, a leading visual content creation and distribution provider, licenses a vast array of visual assets, including photographs, video footage, and illustrations, through its websites to clients worldwide. These assets are central to the lawsuit, as Getty claims that a significant number of these works are original artistic creations where copyright subsists.
The heart of Getty's argument lies in the assertion that Stability AI "scraped" millions of images from its websites to develop and train Stable Diffusion. They contend that the AI's output, in the form of synthetic images accessed by UK users, infringes on their copyright works and trademarks. Stability AI, however, has not yet filed a defense but instead sought to have the case dismissed, focusing on the 'Training and Development Claims' and 'Secondary Infringement Claims' as the central issues of the application.
Significant to the case is the question of whether the training and development of Stable Diffusion occurred in the UK. This issue is crucial as copyright, like database right, is a territorial right confined to the UK's jurisdiction. Getty's claim is based on the allegation that Stable Diffusion was trained using parts of the LAION-5B dataset, which includes millions of images scraped from the web, including those from Getty Images' websites.
The Defendant, Stability AI, asserts that all computing resources used for training Stable Diffusion were located outside the UK, mainly in US data centers operated by Amazon Web Services (AWS). They argue that none of their UK-based employees were involved in the development or training of Stable Diffusion.
However, the court found that there were unanswered questions and inconsistencies relevant to the determination of the location of the training and development of Stable Diffusion. This finding suggests that a fuller investigation might alter the current evidence, underscoring the need for a trial to resolve these issues.
In her judgment, Justice Smith raised serious questions about the credibility of evidence submitted by Emad Mostaque, founder and CEO of Stability AI, in the company's legal battle with Getty Images.
The judge suggested Mostaque's statements may be "inaccurate, or incomplete," noting apparent inconsistencies between his evidence and other available evidence. In particular, she highlighted past interviews where Mostaque discussed fast-tracking visa applications for AI developers from Ukraine and Russia to work in the UK - raising doubts over his claim that no UK-based Stability employees worked on the disputed AI technology, Stable Diffusion.
Justice Smith said the discrepancies cast doubt on the credibility of Mostaque's evidence generally. She ruled that, due to these unanswered questions and possible conflicts in evidence, the issues around Stable Diffusion's development required further investigation at trial.
This lawsuit reflects the broader debate about the use of copyrighted material in training AI models. The outcome could set a precedent for future cases in the rapidly evolving domain of artificial intelligence and its intersection with copyright law.