AI on trial: Fieldfisher and Bird & Bird face off in landmark case

Getty Images’ closely watched copyright claim against Stability AI kicks off at the High Court in London today (9 June), in a case that's being seen as a key test for generative AI in the context of UK copyright law.

The image licensing giant claims the AI company unlawfully scraped millions of its photos to train Stable Diffusion, a text-to-image model used by creators, companies and advertisers. Getty has two main complaints: first, the unlicensed copying of images for training Stable Diffusion's models; and second, that Stable Diffusion’s outputs reproduce substantial parts of its works.

Stability AI, represented by Bird & Bird, denies infringement. Its main argument is expected to be that training took place outside the UK and falls outside the scope of UK copyright law. Getty, advised by Fieldfisher, is also bringing parallel proceedings in the US.

Copyright vs innovation

The case is the UK’s most significant legal test yet on whether training a model on copyrighted material amounts to infringement, and how far UK law can reach. It comes amid a wider global clash between creators and AI developers over data use and ownership.

"The court’s ruling could help decide the future of AI and copyright law in the UK", said Gill Denis, IP senior practice development lawyer at Pinsent Masons. "The decision will test the jurisdictional reach of UK copyright law. It could have a major impact on market practice and the government’s approach to reform."

Ben Milloy, senior associate at Fladgate, noted that the case centres on two alleged breaches. "The first concerns the unlicensed scraping of Getty’s images to train the model", he explained. "Stability appears to accept Getty images were used, but is expected to argue that training happened outside the UK."

"The second", Milloy added, "relates to the synthetic image outputs of Stable Diffusion produced in response to user prompts, which Getty claims reproduce substantial parts of its copyright works.

"In response, we can expect Stability AI to raise a range of arguments relating to the process of its model not in fact using any part of the copyright works (such that there is no 'copying') as well as arguments that – if there is copying – the process falls within certain 'fair dealing' exceptions to copyright law."

What’s at stake

The ruling is expected to follow within months.

A win for Getty could reshape how AI models are trained and trigger a wave of further litigation. A win for Stability could strengthen fair dealing-style arguments and embolden developers to continue training on publicly accessible datasets.