The highly anticipated trial between Getty Images (Getty) and Stability AI (Stability) commenced today in the UK High Court. This case is expected to be a pivotal moment in the ongoing debate over AI and copyright law and represents the most significant legal test yet on whether training a model on copyrighted material amounts to infringement. The case also takes place against the backdrop of debate in UK parliament – where the Commons and Lords have been locked in a similar dispute after the government proposed that copyright holders would have to opt out of their material being used to train algorithms and produce AI-generated content, otherwise it would be free to use by tech companies.

Getty Images, a leading provider of stock photography that syndicates the work of about 50,000 photographers to customers in more than 200 countries, has accused Stability (a London-based AI company) of unauthorised use of millions of copyrighted images to train its AI model, Stable Diffusion. 

Getty’s arguments in the main are twofold: 

1 Stable Diffusion’s use/copying of images required a paid licence, which it did not have; and 

2. that Stable Diffusion’s outputs reproduce substantial parts of its works.

Stability, however, deny the allegations, arguing that its AI training practices took place outside of the UK and thus fall outside of UK copyright law. Getty are bringing a parallel case in the US. 

The trial, expected to last several weeks, will address key legal questions, including whether AI-generated content infringes copyright and whether AI companies can legally train their models using publicly available images. The outcome of the case could have far-reaching implications for the AI industry, shaping future regulations on intellectual property and AI development – most notably the proposed UK AI bill. Peter Kyle, the UK technology secretary, intends to introduce a “comprehensive” AI bill in the next parliamentary session to address concerns about issues including safety and copyright; this has been met with concern about delay of regulation of an industry so clearly in need of direction.  

A victory for Getty here would prompt re-evaluation on how AI models are trained, and victory for Stability would undoubtedly embolden developers to continue training models on publicly available materials – at least for now.