Getty Images Secures Partial Victory Against Stability AI in Landmark UK Copyright and Trademark Dispute

Getty Images and its affiliated companies initiated legal proceedings against Stability AI Limited, a UK-based developer of the Stable Diffusion generative AI model, alleging unauthorized use of Getty’s copyrighted images in training the AI system. The dispute raised multifaceted issues involving copyright infringement, database rights, trademark infringement, and passing off, marking one of the most closely watched cases in the evolving landscape of intellectual property law and artificial intelligence.

The claimants—Getty Images (US) Inc, Getty Images International U.C., Getty Images (UK) Limited, Getty Images Devco UK Limited, iStockphoto LP, and Thomas M. Barwick, Inc—collectively own or license millions of high-quality photographic and artistic images, referred to as the “Visual Assets” or “Copyright Works.” Stability AI’s Stable Diffusion model reportedly incorporated approximately 12.3 million visual assets and associated captions scraped from Getty’s websites and other publicly accessible sources without authorization, according to Getty.

Getty Images initially advanced broad claims including primary and secondary copyright infringement, database right infringement, trademark infringement, and passing off. Central to their argument was that Stability AI unlawfully used Getty’s copyrighted works to train Stable Diffusion, that the AI outputs sometimes reproduced Getty’s images or displayed their trademarked watermarks, and that making the model weights available for download constituted secondary copyright infringement. Model weights, integral to the AI’s learned knowledge, are machine-readable parameters encoding the model’s behavior rather than human-readable source code.

Both parties underscored the case’s significance to their respective industries: Getty representing the creative sector and Stability AI representing AI innovation. Getty Images emphasized that their claims do not seek to hinder AI development but aim to protect creators’ intellectual property rights from exploitation that could threaten the sustainability of creative industries.

During the trial, Getty Images narrowed its claims, dropping allegations that the training occurred in the UK due to lack of evidence and withdrawing primary copyright infringement claims concerning AI-generated outputs. The trial thus focused primarily on secondary copyright infringement related to the distribution of the AI model and trademark infringement claims concerning watermarks visible in some AI-generated images.

In November 2025, Mrs Justice Joanna Smith DBE issued a comprehensive 205-page judgment. The court dismissed Getty’s core copyright infringement claim regarding the use of images in training Stable Diffusion, finding no sufficient evidence that training took place within the UK. The court held that importing and distributing the model in the UK did not constitute dealing in infringing copies.

Regarding secondary copyright infringement, the court rejected Getty’s argument that the model weights themselves constituted infringing copies. The judge concluded that while the model weights were influenced by exposure to Getty’s works during training, they did not store the copyrighted images themselves but rather encoded learned patterns and features. Consequently, the claim for secondary infringement failed.

On trademark infringement, Getty alleged that some AI-generated images displayed watermarks identical or similar to its registered trademarks for GETTY IMAGES and iSTOCK, constituting infringement under the Trade Marks Act 1994. The court found that Stability AI had generated some outputs containing these watermarks within the UK, though such instances were limited and historic, primarily linked to early versions of Stable Diffusion. Stability AI was held liable for this limited trademark infringement.

The court declined to rule on passing off claims, determining that such a judgment would not add materially beyond the trademark findings.

This judgment establishes important legal precedents regarding the applicability of copyright and trademark laws to AI models and intangible elements like model weights. It also exposes challenges within UK intellectual property law in addressing emerging AI technologies, signaling potential areas for legislative and policy development.

Legal observers note that attention will now focus on the UK government’s consultation on copyright and generative AI, particularly debates over proposed text and data mining exceptions that could permit AI companies to scrape copyrighted works unless rights holders opt out. The outcome of this case and ongoing policy discussions will significantly influence the balance between protecting creative industries and fostering AI innovation.