The High Court trial between Getty Images and Stability AI began on Monday 9 June, and will address Getty's claims that Stability AI scraped and copied vast numbers of Getty's images without Getty's consent or payment, in order to train its GenAI image generator, Stable Diffusion.
Getty argues that the outputs (i.e. images) from Stable Diffusion resemble its images, including in some cases the (partial) reproductions of its watermarks, due to the model "memorising" Getty's images. Stability AI counters that these outputs were 'contrived' by specific prompts, rather than typical user behaviour.
Peter Nunn, Partner and Intellectual Property lawyer specialising in AI in Mishcon de Reya's Innovation Department shared his perspective with The Times regarding the landmark trial.
Peter highlighted the challenges Getty faces, as “it seems that the use of those works in the training process took place outside the UK”. He further commented that a victory for Getty could “accelerate a change in copyright law to ensure that the UK does not become an AI backwater, where AI companies are afraid to either develop their models or make them available”.
The court will consider numerous issues, including whether the training of Stability AI's models infringed Getty's IP rights, and if offering Stable Diffusion in the UK constitutes infringement. Stability AI argues that users of its image generator, not it, should be liable for any infringing images produced.
The trial's outcome could influence UK copyright law and AI regulation, amidst government consultations on a 'text and data mining exception' to allow copyright works to be used to train AI models in the UK (subject to rights holders opting out) and ongoing legislative debates.
Read the full article