Getty Image Loses Copyright Infringement Claim Against Stability AI in UK’s First-of-its-Kind Ruling

On 4 November 2025, the UK High Court handed down its highly anticipated landmark judgment in Getty Images (US) Inc & ors v Stability AI Limited [2025] EWHC 2863 (Ch). This was the first UK judgment addressing copyright infringement issues arising from the use and training of generative AI models. 

Case summary

The High Court rejected Getty Images (US) Inc & ors (“Getty”)’s secondary copyright infringement claim in relation to Stability AI Limited (“Stability”)’s scraping of millions of Getty images from its websites for use in its AI model, Stable Diffusion, without authorisation. Stable Diffusion generates images in response to prompts made by text or image and was reportedly trained on over 12 million Getty images scraped from its websites.

Getty argued that, by making the model available in the UK, Stability had imported, possessed and dealt with an “article” that was an infringing copy of Getty’s protected images under the Copyright, Designs and Patents Act 1988 (CDPA). Getty also argued trade mark infringement in respect of images generated by Stable Diffusion that included the Getty watermark. 

Secondary copyright infringement

In respect of the secondary copyright infringement claim, the judge rejected Getty’s claim because, although the Stable Diffusion model is exposed to the copyright-protected works during training, the model does not store the training data itself. As a result, the synthetic images are generated without direct access to the underlying training data, and it could not be proved that the outputs were directly derived from Getty’s images. The judge stated that the model ceased being an ‘infringing copy’ once it no longer contained the ‘copy’ (i.e., the copyrighted works that it was exposed to during training).

Trade mark infringement

Getty was, however, partially successful on the trade mark infringement claim. Stable Diffusion was shown to have historically-generated synthetic images displaying the Getty watermark, which the Court held constituted trade mark infringement in limited circumstances where the outputs were capable of causing customer confusion. The judge held that Stability – having ultimate control over the datasets used to train Stable Diffusion – remained responsible for the trade mark infringement, rather than the user who entered the prompt that generated the infringing image.

Comment

The judgment highlights the challenges of applying UK copyright law in the age of generative AI, with the judge recognising the “very real societal importance” of balancing the interests of the creative industry and the AI sector. This issue has been the subject of increasing debate in recent months, reflecting the growing prevalence of scraping publicly available materials for AI training.

As discussed in our post here, earlier this year, the UK Government launched a consultation on its proposed AI legal framework and the implications for copyright, and emphasised the need to balance the interests of these two sectors. In the United States, the Copyright Office also released a report earlier this year concluding that the use of copyrighted materials to train AI models could constitute prima facie infringement of the right to reproduce such works.

However, the report confirmed that, in contrast to the position in the UK, AI developers in the US can, in fairly broad circumstances, rely on the ‘fair use’ doctrine as a defence, allowing them to continue to use copyrighted materials in this way without authorisation.

In terms of UK application, this judgment indicates that an AI developer could, in principle, commit secondary copyright infringement under UK law by making its model available in the UK if it can be shown that the model copies (for example, by storing or reproducing) copyright-protected images. The Court held that an intangible object can be an “article” for the purposes of the CDPA, and that storage can be by intangible means, such as cloud storage, when establishing copying under the CDPA.

However, the decision is fact-specific, and its implications are limited by several important contextual factors. Getty’s initial claims against Stability included additional causes of action relating to the training and development of Stability’s AI model and associated copyright infringement, which were withdrawn before and during trial.

Getty originally sought, by way of primary copyright infringement, to challenge the use of its images to train Stable Diffusion and certain images generated by the model. These elements were withdrawn because Stable Diffusion was trained overseas, and it could therefore not be established that the relevant infringing acts occurred within the UK. As a result, the judgment does not fully address the breadth of potential issues in this field – particularly the key question of the legality under the CDPA of scraping copyright-protected images for AI training.

Looking to the future, further litigation – whether by way of appeal or new claims by Getty or others – may therefore follow in the coming months. For now, the outcome of this case can be seen as a partial win for AI developers, and rights holders will be keen for further clarity on the UK’s position. 

Finally, in relation to potential policy changes, the UK Government is required to publish its full report on the use of copyright works in AI development by March 2026, which should provide further guidance on how these questions will be addressed. In line with earlier proposals, the UK could choose to follow a similar approach to the EU by introducing an exception to current copyright rules for text and data mining that would apply to the training of AI models.

This mechanism would permit copyrighted materials to be used for such purposes unless the rights owner has expressly opted-out. This proposal was publicly criticised by various high-profile creative rights holders when it was first announced. However, the UK copyright framework has remained largely unchanged since the CDPA was enacted in 1988, and this judgment – together with the government’s ongoing engagement – demonstrates the pressing need to reassess and update the framework to reflect significant technological advances.

Subscribe to Ropes & Gray Viewpoints by topic here.