In one of the first AI-related judgments in the UK, the High Court refused Stability AI’s (Stability) application for reverse summary judgment against Getty Images (US) Inc and its related group companies (Getty).
The proceedings concerned allegations that Stability (a generative AI company) had scraped millions of images from Getty’s website without consent and unlawfully fed the scraped images into its ‘Stable Diffusion’ AI model. Getty alleged that Stable Diffusion’s output (images created based on a text description) infringed Getty’s copyright by closely replicating Getty’s visual assets and replicating its trademarks (together with other allegations for passing off and data base infringement).
Stability sought reverse summary judgment and / or strike-out of two elements of Getty’s claim: firstly, in relation to the claim for copyright and database infringement on the grounds that Getty’s visual assets were downloaded by Stability during the training and development of Stable Diffusion (the Training and Development Claim); secondly, in relation to the claim for so-called ‘secondary infringement’ on the basis that Stability imported Stable Diffusion software into the UK (the Secondary Infringement Claim).
The Training and Development Claim
Arguments in respect of the Training and Development Claim focused on the evidence as to whether Stable Diffusion was trained and developed in the UK. This is a vital element for claims relating to copyright and database rights, as these are territorial rights, meaning that for the Training and Development Claim to ultimately succeed, Getty would have to prove that Stable Diffusion was trained and developed within the UK.
Getty’s original claim relied on the location of Stability’s employees to infer that Stable Diffusion was trained within the UK. Stability’s reverse summary judgment / strike out application argued that none of its UK employees ever worked on Stable Diffusion and that all its training infrastructure was based in the US. The Court, however, found that there remained a reasonable prospect that Getty would be able to refute this point, partly on the basis of new evidence put forward by Getty.
The Court noted that Stability did not adequately address Getty’s new evidence in its own witness evidence in support of its application and also considered that Stability did not conduct sufficiently thorough investigations of UK employees’ roles in relation to Stable Diffusion. Similarly, the Court found that questions around the type of software used to train Stable Diffusion, which went to the heart of where the model was trained, were not adequately addressed by Stability in its application. Ultimately, these points left the Court unable to determine that Getty’s claim had no reasonable chance of success.
The Secondary Infringement Claim
Arguments in respect of the Secondary Infringement Claim centered on the correct interpretation of the word ‘article’ in sections 22, 23 and 27 of the Copyright, Designs and Patent Act 1988 (CDPA). Broadly speaking, these sections make it an offence to import into the UK, sell or use in business copyrighted ‘articles’.
Getty alleged that ‘articles’ can be intangible by nature, meaning that, with Stable Diffusion being available in the UK, Stability was in fact importing an ‘article’ into the UK in breach of section 22 CDPA.
The Court suggested that the question of whether an ‘article’ could be something intangible remains unresolved by English courts and the Judge was therefore hesitant to opine on this point of law in the context of a summary judgment hearing. The Judge considered that it would be inappropriate for the Court to hand down a judgment in such a novel area of law without properly assessing the facts in the context of a full trial.
What now?
The case, including Getty’s claims which were not the subject of this application, will proceed to a full trial, subject to any settlement agreement or further interim applications.
Observations
- Unusually, each party relied on a substantial body of factual and expert evidence at the summary judgment hearing. This demonstrates the factual and technical complexity of AI-related cases, and perhaps indicates that (for now at least) these claims may not be suitable for disposal via summary judgment.
- In addition, questions by the Court about the expertise of Getty’s expert witness reinforce the need for expertise tailored to the issues in dispute. In nascent industries such as AI, such expertise is likely to be even more scarce, underlining the need for early identification and engagement with potential experts.
- Finally, the Court did not consider that parallel claims in a different jurisdiction were a barrier to winning the “territory” point in relation to copyright and database right claims. Rather, the Court recognised the possibility that training and development of AI tools may take place across multiple jurisdictions.
