IL-2023-000007 - [2025] EWHC 2863 (Ch)

INTRODUCTION

These are proceedings alleging primary and secondary copyright infringement, database right infringement, trade mark infringement and passing off against the Defendant (“Stability”), an open-source generative artificial intelligence (“AI”) company. The claim concerns Stability’s deep learning AI model (known as “Stable Diffusion” or “the Model”).

Shortly prior to closing submissions, the Claimants (collectively referred to as “Getty Images”) abandoned various aspects of their claim, thereby narrowing the issues to be determined by the court and rendering large parts of the opening submissions and evidence irrelevant. Nevertheless, the claim continues to raise issues of importance in the field of intellectual property in connection with the use of AI models such as Stable Diffusion.

Deep learning AI models use an ‘artificial neural network’ architecture designed to approximate the structure of synaptic connections in the brain. The neural network consists of multiple layers (hence the term deep) which create a hierarchical processing structure.

Stable Diffusion is a type of generative AI model known as a diffusion model, or more specifically a latent diffusion model. Broadly, it transforms an input (a user command or “prompt” in the form of written text or a “seed” image) into a desired output in the form of a synthesised image by modelling a probability distribution based on its training data and then sampling from that distribution. The development of a stochastic model of this type typically involves designing and building the architecture for the model which is then trained by repeated exposure to massive quantities of data, in this case in the form of human-generated digital images contained in datasets created by crawling and scraping images and associated descriptive captions from the Internet.

The model parameters (the “model weights” or “biases”) define the network connections in the model and are learnable parameters controlling the functionality of the network. Before training begins, the network’s weights are initialized with random values. As the network is exposed to the training data, the weights are iteratively updated to better satisfy an optimization criterion specified by engineers. Once the model is trained, running the network, referred to as inference, is (in simple terms) an input-output system in which the user specifies inputs, the trained network performs computations on those inputs and then generates the desired output.

Although there are differences between the various versions of Stable Diffusion, essential to these models is a process which starts with a random noise image. The trained network (conditioned by the user-specified prompt) iteratively removes the noise so as to create an image which is semantically consistent with the prompt.

Inference does not require the use of any training data and the model itself does not store training data. However, a large part of its functionality is indirectly controlled via the training data. In other words, the way in which the network makes use of its multiple layers is the result of the training process.

The claim as pleaded seeks to protect what Getty Images describe as “the lifeblood” of their business, namely millions of high-quality photographic images of, amongst other things, world events, sporting moments, celebrities, architecture, nature and travel (“the Visual Assets”), created over many years by hundreds of thousands of photographers. At the heart of that case is the allegation that, in order to develop and train Stable Diffusion, Stability has scraped millions of Visual Assets (a substantial proportion of which are said to comprise original artistic works and/or film works owned by, or exclusively licensed to, the First Claimant in which copyright subsists (“the Copyright Works”) from Getty Images’ websites without consent and used those images unlawfully to train and develop a number of versions of Stable Diffusion.

Notwithstanding their pleaded case, it is now acknowledged by Getty Images that (i) there is no evidence that the training and development of Stable Diffusion took place in the United Kingdom (such that what has been called “the Training and Development Claim” has been abandoned); (ii) the prompts which it was alleged had been used to generate the examples of infringing output from the Model in evidence in these proceedings have been blocked by Stability such that the relief to which Getty Images would have been entitled in respect of their allegations of primary infringement of copyright (referred to as “the Outputs Claim”) has now been substantially achieved. Thus the Outputs Claim has also been abandoned; and (iii) given its inherent link to the Training and Development Claim and the Outputs Claim, a claim for database rights infringement (“the Database Rights Infringement Claim”) can now no longer be advanced.

However, Getty Images continue to advance a case that normal use of Stable Diffusion by users in the United Kingdom will, in some cases, generate synthetic images bearing Getty Images’ own trade marks, contrary to sections 10(1), 10(2) and 10(3) of the Trade Marks Act 1994 (the “TMA”) (“the Trade Mark Infringement Claim”), and/or constituting an actionable misrepresentation under the law of passing off (the “Passing Off Claim”). They also maintain a case that, contrary to sections 22 and 23 of the Copyright, Designs and Patents Act 1988 (the “CDPA”), Stability has imported into the UK, otherwise than for private and domestic use, possessed in the course of business, sold or let for hire or offered or exposed for sale or hire, or distributed an article, namely Stable Diffusion, which is and which Stability knew or had reason to believe is an infringing copy of the Copyright Works.

Getty Images do not say that Stable Diffusion is itself a copy of, or that it stores within it any copies of, the Copyright Works. However, pursuant to section 27(3) CDPA, Getty Images contend that Stable Diffusion is an infringing copy under the CDPA because the making of its model weights would have constituted infringement of the Copyright Works had it been carried out in the UK (“the Secondary Infringement Claim”).

Both sides emphasise the significance of this case to the different industries they represent: the creative industry on one side and the AI industry and innovators on the other. Where the balance should be struck between the interests of these opposing factions is of very real societal importance. Getty Images deny that their claim represents a threat to the AI industry or an attempt to curtail the development and use of AI models such as Stable Diffusion. However, their case remains that if creative industries are exploited by innovators such as Stability without regard to the efforts and intellectual property rights of creators, then such exploitation will pose an existential threat to those creative industries for generations to come.

As to whether this judgment will, in reality, have anything to say on the balance to be struck between the two warring factions, it is worth observing at the outset that this court can only determine the issues that arise on the (diminished) case that remains before it having regard to the available evidence and the arguments advanced by the parties. It is no part of this court’s task to consider issues that have been abandoned or to consider arguments that are no longer of relevance to the outstanding issues.

Attached to this Judgment at Appendix A is a Glossary of the key technical terms used in this Judgment when discussing the technology. Those terms have also been emboldened in the text of the Judgment for ease of reference.