Getty Images v Stability AI - UK trial begins... (Part 1)

Yesterday marked the first day in the highly anticipated Getty Images (Getty) v Stability AI (Stability) trial before the High Court in London. Here we cover the opening salvos and the trade mark infringement and passing off claims. Tomorrow, we will pick up the copyright and database rights claims, and safe harbour defence.

The trial is set to last for 18 days, and there were, as of the last reported court judgment in the case, 63 issues for determination. In that judgment, the judge noted that the “issues arising have required the court to devote unprecedented amounts of court time to dealing with various disputed issues that have already arisen”, and that there had been four hearings this year in as many months, with a fifth (the pre-trial review) having been listed for late May.

The case started in 2023 (see Guest Post from this Kat and Katfriend, Adrian Aronsson-Storrier here). The parties have been liaising over a ‘technical primer’ but it appears there has even been a dispute about that. Helpfully the IPKat has you covered with a primer provided by Katfriend Jakub Wyczik here. The case has resulted in three written judgments already, starting back in December 2023, with applications for summary judgment and amendments to the statements of case ([2023] EWHC 3090 (Ch)), a judgment on whether one of the Getty claimants could act as a representative for over 50,000 contributors/licensors ([2025] EWHC 38 (Ch)), and pre-trial application to introduce what was called a ‘statement of case on development and training’ to supplement its particulars of claim ([2025] EWHC 109 (Ch)). 

It is fair to say that the case is both legally and factually an absolute behemoth, including allegations of trade mark infringement, passing off, copyright infringement, and database right infringement. Even that summary underplays the complexity of the issues; the copyright claim itself addresses knotty issues including communication to the public, the requirements for having authorised users to infringe copyright, safe harbours under the UK’s transposition of the E-Commerce Directive, and novel questions of secondary infringement. And to make matters even more complex, there are complex issues around ownership and licensing, and the overarching context that it concerns the training, development and deployment of AI.

It is a very hard-fought case. The skeleton arguments run, between them, to 560 pages. Getty says in its skeleton that “[Stability] has ridden roughshod over this vast creative and entrepreneurial endeavour by simply taking millions of images, which were scraped from Getty Images’ websites without its consent, and using them as key input to train its commercial AI model, Stable Diffusion, the output of which competes with Getty Images’ content.” Stability’s skeleton notably claims that Getty’s claims in relation to outputs “are so weak that their introduction seems likely the result of an intuitive reaction to Stable Diffusion and not a clear-headed analysis of the rights in issue.” Meow.

Keeping track of the myriad of issues must be a challenge for the legal teams, let alone those of us watching from the terraces. However, I will endeavour in this post and the next to distil the issues that will be of interest to readers as follows.

Trade mark infringement

This issue revolves around Getty’s case that images produced by Stable Diffusion have sometimes included the Getty Images watermark. Here is an example:

Getty’s case is that those watermarked images have been produced both to Getty, its legal team and its expert, as well as to third parties. It claims that this is a function of the volume of Getty’s images that have been included in the training data set; on Getty’s case, it is millions of watermarked images.

Stability’s case is that those watermarks were “contrived” with atypical prompts designed to create the responses sought, and that even if those watermarks did appear to typical users, there is no use in the course of trade. Stability argues that it is the user that generates any signs identical/similar to the trade marks, Stability cannot control that, and that the user seeing the image would not view the sign as being a commercial communication from Stability.

This claim has reportedly led to an early skirmish at trial, with Rebecca Newman (Addleshaw Goddard) reporting on LinkedIn that the judge refused to permit Getty to rely on the possible generation of child sexual abuse material (as part of its tarnishment claim) on the basis that it was not properly pleaded, to which Getty sought permission to appeal and suggested that the trial may need to be adjourned pending an application. Permission was refused, leaving the trial continuing, albeit with the spectre remaining of the Court of Appeal derailing matters.

The trade mark claim is an unusual one. The appearance of watermarks is a glitch rather than a feature; users would not typically want a watermark in an image. However, there could be some interesting consideration of cases (such as Louboutin ECLI:EU:C:2022:1016) as to whether the appearance of the watermarks is ‘use’, and on the origin function. There will also be some interesting evidential points for the court to consider about whether the evidence obtained by Getty’s lawyers/experts’ use of Stable Diffusion was done with the consent of Getty. The best source of evidence Getty could obtain is by it and its representatives using Stable Diffusion, which Stability says must have been done with consent.

Passing off

To non-UK readers, passing off is the UK’s closest equivalent to unfair competition (aside from unfair advantage (etc.) for registered trade marks). Stability has admitted that the Getty Images marks have goodwill, but denied that there is any misrepresentation or damage to that goodwill. It is unclear to this Kat whether the passing off claim really adds much to the trade mark claims.

To be continued...

Tomorrow, we will cover the copyright and database rights claims. We will also cover the safe-harbour defence regulations 18 and 19 of the Electronic Commerce (EC Directive) Regulations 2002 (derived from European Directive 2000/31/EC).