- within Family and Matrimonial, Insolvency/Bankruptcy/Re-Structuring and Strategy topic(s)
In an important decision considering the interplay between Artificial Intelligence ("AI") and English copyright and trademark laws, the English High Court handed down judgment in the longstanding battle between visual content creator, Getty Images and AI developer, Stability AI on 4 November 2025 (the "Judgment").1 Mrs. Justice Joanna Smith DBE handed Getty Images a pyrrhic victory, ruling that it was partially successful on its trademark infringement claim, with the scope of the Court's findings said to be "both historic and extremely limited". The Court otherwise dismissed Getty Images' passing off and secondary copyright infringement claims.
Although much more limited in scope than was once expected, the decision provides helpful guidance on how the Courts will apply certain existing intellectual property law to generative AI models. Given the speed with which AI is developing and the huge volumes of data being used to train AI models, the need to balance the freedom of AI developers with the rights of content creators has become a hot topic both in the UK and more globally. It has prompted the English Government to consult on how it can deliver a solution which ensures that both developers and content creators can share in the benefits of AI, with the Government not ruling out the need for direct intervention through legislative change.2 The decision in Getty Images is therefore an important first step in understanding whether, from an English law perspective, there is a need for Parliament to clarify the law in this area, with the Judge recognising the "very real societal importance" of finding a balance between the interests of the creative industry and AI developers.3
Against that background, this article looks at Getty Images' claims against Stability AI, the Court's decision on the claims that survived through to the end of trial, as well as looking at two ongoing lawsuits brought against Stability AI in the United States: a parallel action between Getty Images and Stability AI and a class action against Stability AI and other AI companies
Getty Images v. Stability AI: English Claim
Getty Images is a leading creator and distributor of digital content. It owns a vast online collection of photographic images and various trademarks relating to the Getty Images and iStock name and logo. In January 2023, Getty Images commenced proceedings in the English High Court against Stability AI, the developer of Stable Diffusion, which is an open-source, deep learning AI model that generates images from user text prompts. Stable Diffusion was trained using online images, including from Getty Images' websites, and it was alleged that the use of this content infringed Getty Images' copyright under the Copyright, Designs and Patents Act 1988 (the "CDPA"). Getty Images also asserted trademark infringement and passing off claims.
Primary Copyright Infringement (Abandoned During Trial)
Getty Images raised primary copyright infringement on two bases: (1) Stability AI's use of Getty Images' content to train and develop Stable Diffusion, and (2) the reproduction of Getty Images' content in synthetic image outputs of Stable Diffusion.4
- Training and Development Claim: Getty Images alleged that Stability AI "scraped" millions of images from Getty Images' websites without consent and used those images unlawfully as input to train and develop Stable Diffusion. Although Stability AI admitted that at least some images were used during the training of Stable Diffusion, it disputed that the relevant training activities, including the downloading and storage of images on servers, occurred in the UK. The question of where those training activities took place was the subject of a reverse summary judgment application heard in late 2023. Although the Judge dismissed the application, noting that it was not for the Court to conduct a mini-trial on an application of that sort,5 she held that the evidence before the Court would (on its face) provide "strong support"6 for a finding that no development or training of Stable Diffusion had taken place in the UK. Getty Images' case did not materially improve following disclosure, and a late application to amend its "inchoate inferential"7 case that certain development and training activities took place in the UK was refused two months before trial. Moreover, during trial, Stability AI's witnesses testified that no training datasets were downloaded onto UK computers or servers,8 making Getty Images' primary copyright infringement case significantly harder to establish. It was therefore unsurprising that Getty Images elected to drop this part of its claim during trial.
- Output Claim: Getty Images also abandoned its claim that the synthetic images produced as an output of Stable Diffusion reproduced a substantial part of Getty Images' works. There are a number of factors which could have influenced that decision, including the fact that by the time trial began: (1) Getty Images was relying on only 13 representative examples of works whose copyright was said to have been infringed (a number that was further reduced during trial); (2) title to those images was still in dispute; (3) Stability AI had subsequently blocked the ability to generate those particular copyrighted images in Stable Diffusion; and (4) the claim was largely unsupported by the expert and factual witness evidence.
Secondary Copyright Infringement
In light of the abandonment of its primary copyright infringement claim, Getty Images was left to rely on its claim for secondary copyright infringement under sections 22 and 23 of the CDPA. That claim focused on the allegation that the AI model itself, Stable Diffusion, was an "article" imported into the UK which was an "infringing copy" of Getty Images' copyrighted works. This gave rise to two important legal questions, namely whether a deep learning AI model is capable of being (1) an "article" for the purposes of sections 22 and 23 of the CDPA, and (2) an "infringing copy" as defined in section 27 of the CDPA. Those questions not only turned on the statutory construction of the CDPA but also the evidence given at trial as to how the AI model actually works.
As to construction, the Judge was clear that the words of the statute must be construed having regard to their proper context, including advancements in technology. In her view, an electronic copy stored on an intangible medium is capable of being both an "infringing copy" and an "article" within the meaning of sections 22, 23 and 27 of the CDPA.9 However, in order to establish that there was an "infringing copy", it was necessary to show that there had, in fact, been a reproduction of copyrighted works. In this case, the issue turned on whether an AI model, such as Stable Diffusion, which results from a training process involving the exposure of model weights to infringing copies, but which never actually contains or stores those copies, is itself an "infringing copy". On the facts of this case, the Judge determined that it was not.10
Trademark Infringement and Passing Off
Getty Images' trademark infringement and passing off claims rested on the allegation that certain image outputs of Stable Diffusion contained Getty Images and iStock watermarks, contrary to sections 10(1)-(3) of the Trade Marks Act 1994 ("TMA"), or alternatively, that such acts constituted actionable misrepresentations under the law of passing off. Getty Images argued that the existence of those watermarks on the outputs misrepresent the origination of the images and Getty Images' endorsement.
For the trademark infringement claim to succeed, Getty Images had to establish that Stable Diffusion generated at least one output in the UK which contained the protected watermarks. After an exhaustive consideration of the evidence – including the results of experiments undertaken by the experts as to the generation of synthetic images and evidence of watermark generation "in the wild"11 – the Court found, on balance, that this threshold question had been met in relation to two of the earliest versions of the AI model.12 And in comparing the watermarked images generated by Stable Diffusion with the Getty Images/iStock watermarks, the Court concluded that sections 10(1) and 10(2) of the TMA had been infringed in an "extremely small number of examples".13 Specifically, the Judge made a finding under section 10(1) (double identity) in relation to only one version of Stable Diffusion and in respect of the iStock watermark only; and a finding under section 10(2) (likelihood of confusion) in relation to certain outputs generated in two versions of the model (one in respect of iStock watermarks and the other concerning Getty Images watermarks).14 However, the Judge had no means of assessing the number of actually infringing watermarks in the real world and Getty Images did not run any case based on the probability of infringing watermarks being produced by those (early) versions of Stable Diffusion. The findings were therefore both "historic and extremely limited in scope".15 The Court otherwise rejected Getty Images' claim under section 10(3) that the use of the watermarks amounted to an unfair advantage, detriment to distinctive character and/or damage to reputation.16
As for the passing off claim, Getty Images' position at trial had been that this claim would "stand or fall" with the trademark infringement claim. Whilst the Judge invited the parties to make further submissions on this issue in light of her findings on trademark infringement, neither did so and, as such, the Court declined to consider the point further.17
Getty Images v. Stability AI: US Dispute
In February 2023, Getty Images initiated a parallel suit against Stability AI in the US District Court for the District of Delaware.18 Shortly after filing, Stability AI asked the Court to dismiss the case due to lack of jurisdiction, or alternatively, to transfer it to another District Court in California. In July 2024, following a year of narrow discovery limited to jurisdiction issues, Stability AI renewed its motion to dismiss. However, in the absence of the Delaware Court taking any action in relation to that motion, Getty Images voluntarily dismissed the case on 14 August 2025 and filed another complaint that same day in the US District Court for the District of Northern California, where Stability AI had hoped to transfer the prior case.19
The complaint Getty Images filed in California asserts claims similar to those in the UK action, and those in the Delaware action prior to dismissal. Getty Images alleges copyright infringement under the US Copyright Act based on the training and development of Stable Diffusion, as well as output images generated by Stable Diffusion. In connection with the display of Getty Images' watermarks on output images, Getty Images asserts a claim for providing false copyright management information. Getty Images also asserts various causes of action relating to trademark infringement, trademark dilution, and unfair competition under the US Lanham Act and California state law. The California Court has not yet entered any substantive rulings on Getty Images' claims, and it remains to be seen what impact (if any) the English Court's Judgment will have on Getty Images' appetite to pursue the US action.
Andersen v. Stability AI: US Class Action
In addition to defending proceedings brought by Getty Images on both sides of the Atlantic, Stability AI (along with certain other AI companies) are facing a class action lawsuit brought in California in January 2023 by a number of visual artists for copyright infringement relating to the training and development of Stable Diffusion.20 This was not the first group action that Stability AI has faced, having already successfully challenged a representative action brought in the UK on behalf of around 50,000 photographers and content creators for copyright infringement.21
The US class action has gone through a series of motions to dismiss and, unlike the position in England, the US Court has ultimately allowed certain of the allegations to proceed to trial. This includes the direct copyright infringement claims based on the defendants' alleged copying and use of "Training Images" in their respective AI models.22 The Court also declined to dismiss direct copyright infringement claims that were based on two other theories, namely (1) the "Model Theory" (which is described as the theory that "the Stable Diffusion 1.5 product itself – after it was trained – is 'an infringing Statutory Copy' of plaintiffs' works or a 'Statutory Derivative Work' because it represents a transformation of plaintiffs' works"); and (2) the "Distribution Theory" (which is based on allegations that the defendants' infringed "plaintiffs' exclusive distribution rights because distributing Stable Diffusion 1.5 is equivalent to distributing plaintiffs' works").23 The Court noted that both theories depend on whether the protected works are "contained, in some manner, in Stable Diffusion", and further noted that whether the works are contained as algorithmic or mathematical representations "is not an impediment to the claim at this juncture".24
Additionally, the California Court refused to dismiss the induced copyright infringement claim asserted against Stability AI.25 Stability AI had argued there were no allegations that it "promoted the use of Stable Diffusion to 'infringe copyright,'" but the Court pointed out, "The plausible inferences at this juncture are that Stable Diffusion by operation by end users creates copyright infringement and was created to facilitate that infringement by design".26
Trial for the case is currently set for April 2027 and it will be interesting to see whether between now and the start of trial plaintiffs try to distance themselves from any legal or factual findings in the Judgment which could give rise to issue preclusion or collateral estoppel (i.e., the doctrine that prevents facts or legal determinations from being re-litigated in separate proceedings).
Conclusion and Takeaways
Although the Judgment was eagerly awaited, the fact that Getty Images' claims largely failed was not entirely surprising following the developments at trial. Insufficient evidence and jurisdictional issues significantly impacted the scope of the Court's findings, leaving unanswered the primary copyright infringement claim concerning the training, development, and operation of the AI model. What the decision does show is the challenges Courts face in applying existing intellectual property laws to modern and continually evolving technology, an issue which the UK Government considers is holding back growth for the AI and creative industries.27sup> For the time being, and based on the Judgment, AI developers may feel that they can avoid infringing English copyright law simply by ensuring that the AI model does not itself store or reproduce data in the UK. Whether content creators can successfully pursue AI developers for copyright infringement in other jurisdictions remains to be seen but, for Getty Images, this will have been a bruising experience before the English Courts that it will not wish to repeat when its parallel claim comes to be heard by the California Court in due course.
Footnotes
1 Getty Images (US) Inc & Ors v Stability AI Ltd [2025] EWHC 38 (Ch). The claim was brought by various entities within the Getty Images group, together with a US company which is the trading vehicle of the photographer, Thomas Barwick.
3 Judgment §§12-13.
4 Given that both of these claims were abandoned during trial, the Judge did not address them in the Judgment.
5 Getty Images (US), Inc & Ors v Stability AI Ltd [2023] EWHC 3090 (Ch), at §62.
7 Getty Images (US), Inc & Ors v Stability AI Ltd [2025] EWHC 109 (Ch), at §29.
8 Ibid, at §100
9 Judgment §§574-583.
10 Judgment §§592-602.
11 Judgment §§144-231.
12 Judgment §257.
13 Judgment §496.
14 Judgment §§315-453 (section 10(1)) and §§454-467 (section 10(2)).
15 Judgment §757. See too §496.
16 Judgment §§468-538.
17 Judgment §§539-542.
18 Getty Images (US), Inc. v. Stability AI, Inc., No. 1:23-cv-00135-JLH (D. Del.).
19 Getty Images (US), Inc. v. Stability AI, Ltd., No. 3:25-cv-06891-TLT (N.D. Cal.).
20 Andersen v. Stability AI Ltd., No. 3:23-cv-00201-WHO (N.D. Cal.).
21 The English Court refused to allow the representative action to continue, primarily on the basis that it was not satisfied the class could be adequately defined or that the representative claim would avoid the need for an individualised assessment of the liability and quantum issues: Getty Images (US) Inc & Others v Stability AI Ltd [2025] EWHC 38 (Ch), at §83.
22 Andersen v. Stability AI Ltd., 700 F. Supp. 3d 853, 864 (N.D. Cal. 2023).
23 Andersen v. Stability AI Ltd., 744 F. Supp. 3d 956, 974 (N.D. Cal. 2024)
24 Id.
25 Id. at 968-69.
26 Id. (citation omitted).
27 Supra note 2.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
