Introduction
The meeting point of artificial intelligence and intellectual property law has become one of the critical legal questions of the digital age. At the centre of that debate, currently, sits Getty Images (US) Inc & Ors v Stability AI Limited [2025] EWHC 2863 (Ch), a case that arrived with enormous promise for rights holders and concluded in November 2025 with a ruling that was, at best, partial and inconclusive.
Getty Images is a stock media library selling footage and still images on behalf of photographers, offering rights managed, royalty free and stock media licensing. The global stock library industry is purported to be worth around US$6 billion and rising, supporting a large community of creatives who depend on it for their income. The rise of AI image generation poses a direct threat to that ecosystem, which is what made Getty’s decision to sue Stability AI in the English courts in January 2023 so significant.
For those working in the art market, publishing, photography and digital media, the stakes are therefore profound. The case tests whether generative AI can lawfully train on protected creative works and, if so, on what terms.
When Art Meets Algorithm, Copyright and Artificial Intelligence: The UK High Court Decision
Getty Images, one of the world’s largest visual content libraries representing more than 600,000 photographers and artists, issued proceedings in the English High Court in January 2023. It alleged that Stability AI had scraped over 12 million Getty images without authorisation to train its generative AI model, Stable Diffusion.
Getty’s claim was extensive. It included primary copyright infringement, secondary copyright infringement, database right infringement, trademark infringement and passing off.
By the time trial began in June 2025, the scope had narrowed dramatically. Getty conceded that the training and development of Stable Diffusion took place outside the United Kingdom. That concession proved decisive. It removed the court’s jurisdiction over primary copyright infringement, leaving only secondary infringement and trademark issues to be determined.
AI Training and Copyright Infringement, What the Courts Have Decided So Far
On 4 November 2025, High Court Judge Joanna Smith DBE handed down judgment in Getty Images v Stability AI. On copyright, Getty lost. The court held that Stable Diffusion’s model weights were not an “infringing copy” within the meaning of the Copyright, Designs and Patents Act 1988. (The model weight being a numerical value, a number, used in determining how Images and text fed into the model are trained and interpreted for the output/outcome).
The reasoning was technically significant. Model weights are not stored reproductions of images. Rather, they are statistical parameters guiding the outcome of a learning process. As the court explained, the weights are “purely the product of the patterns and features which they have learnt over time during the training process through the process of gradient descent.” In short, the model learns patterns rather than retaining images. Thus, Stability AI’s central defence held.
However, the court made an important finding in principle. It accepted that an intangible electronic copy can constitute an “article” for the purposes of secondary infringement. That clarification may assist future claimants, even though it did not help Getty on the facts of this case.
Getty achieved a limited success on trademark infringement. The court found that certain early versions of Stable Diffusion generated outputs displaying the Getty watermark. That amounted to infringement under sections 10(1) and 10(2) of the Trademarks Act 1994. It was a symbolic win rather than a sweeping one.
Crucially, the central question remained unanswered. The court did not determine whether training an AI model on copyrighted works without a licence constitutes infringement under UK law. That issue was left unresolved.