Navigating representative actions: takeaways from Getty Images v Stability AI

On 14 January 2025, the High Court delivered a significant judgment in Getty Images (US) Inc and others v Stability AI Ltd. The case is already shaping up to be a landmark case, addressing the complexities of the application of intellectual property rights in the digital era, in particular alleged infringement arising from the use of copyright materials in the training of artificial intelligence (AI) models. However, this instalment focuses on the role of the Sixth Claimant, a named individual, as a representative of a large number of copyright owners, an issue which goes to the heart of how the law grapples with mass claims in the "big data" era, where claims centre around the use of the data and digital works of vast swathes of people en-masse in the development of tech products such as large language AI models.

Background of the Case

The claimants (including Getty Images (US) Inc, Getty Images International UC, Getty Images (UK) Limited, Getty Images Devco UK Limited, iStockphoto LP (the "Getty Images Group") and Thomas M Barwick Inc (the "Sixth Claimant")) brought a lawsuit against the defendant, Stability AI Ltd – an open-source generative AI company based in England and Wales, alleging that Stability AI Ltd. infringed their intellectual property rights by using their copyrighted images without authorisation, to train the AI model, Stable Diffusion. The claims encompassed copyright infringement, database right infringement, trademark infringement, and passing off.

The core of the dispute revolves around Stability AI’s use of datasets containing images taken without consent from the Getty Images Websites to train 'Stable Diffusion'. Stability AI, admits that "at least some images from the Getty Images websites were used during the training of Stable Diffusion", but has not identified those specific images. The claimants argue that AI-generated outputs reproduced substantial parts of the scraped images.

A key part of the claim involved the Sixth Claimant, who sought to act as a representative of a class of copyright owners, who have exclusively licensed their works to the Getty Images Group and which were said to have been scraped and used in the training of Stable Diffusion.

Copyright Infringement Claims

The claimants' copyright infringement claims are:

1. The defendant downloaded, on servers or computers in the United Kingdom, and used, their copyrighted works during the training of Stable Diffusion. This claim alleges direct infringement of the copyright by the defendant, by directly engaging in the restricted rights of the owner of the copyrighted work set out in Section 16 of the Copyright, Designs and Patents Act 1988 (the "CDPA'). Acts of primary infringement do not require knowledge or intention – they are "strict liability".
2. The defendant imported the pre-trained Stable Diffusion software into the UK, which contained infringing content. This second claim alleges a secondary infringement pursuant to section 22 of the CDPA for importing infringing copies. Unlike direct infringement, secondary infringement requires some knowledge that the import was an infringing copy of a copyright work.   
3. Finally, the synthetic images generated by Stable Diffusion are also said to be infringing since they reproduced substantial parts of the claimants' copyrighted works.

Representative Claim

Issue

The claimants' action included a representative action on behalf of over 50,000 copyright owners whose works were exclusively licensed to the Getty Images Group and included works which were used to train Stable Diffusion. They argued that each of these copyright owners had a concurrent right of action alongside the Getty Images Group through the Sixth Claimant, who “represents and has the same interest in this claim as the parties who are owners of artistic works and film works that have been licensed on an exclusive basis to the First Claimant”. The claimants further argued that despite variations in the contributor agreement, the relevant content was consistently licensed exclusively to the Getty Images Group, allowing them to bring infringement proceedings, and any diversity of interest among the represented parties could be resolved during the inquiry into damages or profits.

The defendant issued an application to challenge the representative claim of the Sixth Claimant, asserting that individual copyright owners had different agreements and levels of exclusivity, necessitating individual assessments rather than a single, collective claim.

Court's Decision

The representative claim raised multiple complex issues. The court identified two main questions to be considered in relation to the representative claim:

1. is this case within CPR 19.8 which allows representative actions when multiple parties have the "same interest" in a claim ?¹ and
2. if so, nevertheless should the court in its discretion direct that the Sixth Claimant may not act as a representative?

The court determined that it did not have jurisdiction to permit the claim to proceed as a representative action, or alternatively that it should not permit the claim to proceed in the exercise of the court's discretion. The key difficulty was the class definition, which depended (in part) on whether a copyright owner's works included works which were used to train Stable Diffusion.  There was no definitive list of the copyright works that had been used to train Stable Diffusion and the defendant had only made limited admissions to the effect that “at least some” had been used. Therefore it was not possible to say of any particular person whether or not they qualified for membership of the represented class. ²  The court rejected the claimant's argument that this was in fact a matter within the defendant’s own knowledge. It therefore refused to allow the representative claim to proceed, saying it could not be satisfied that the representative claim would remove the need for an expensive and time consuming individualised assessment of numerous issues of liability and quantum relating to the proposed represented parties, and also citing the potentially very significant case management burden for the court.

Sampling and Extrapolation

The court acknowledged that a sampling regime, along with cooperation between the parties, could potentially help manage the litigation more effectively. However, the lack of clarity on how such sampling would be implemented and whether it is even feasible made it an insufficient reason to permit a representative claim. The judgment emphasised that whilst large-scale copyright claims in AI-generated content present unique challenges, they require structured, practical legal mechanisms rather than overly broad representative actions.

CPR 19.3 Application

The Claimants additionally argued that, if the court refused a representative claim, justice would require permission under CPR 19.3(1) for the copyright owners to bring their own claims. They also claimed that the Right To Control Claims Clause would protect the defendant from future claims.

The defendant countered  that the application was raised too late and did not assure protection from future claims. They argued that using CPR 19.3 as a fallback for a failed representative claim under CPR 19.8 would be an error of principle.

The court refused the claimants' informal application under CPR 19.3 concluding that:

• There was insufficient evidence to support the application.
• Allowing such an application would be unfair to the defendant given the risk of the enormous potential further claims by copyright owners.
• A key consideration was the diversity in licence agreements between the Getty Images Group and copyright owners, some of which contained Right To Control Claims Clauses in different forms, while others lacked them entirely.

Conclusion

Although the representative claim was not allowed to continue, the Judge emphasized the need for a pragmatic solution, that avoids joining 50,000 potential claimants.  One of the issues was the lateness at which this issue was coming into focus, given all the other matters involved in the case.  The Judge was critical of the defendants for not challenging the way the representative action was put together sooner, but offered suggestions for how the matter of these complex claim may be addressed in future.

The Judge suggested that an order under CPR 19.3 could make sense, as it would allow the issues to be dealt with using sample cases and focusing on the "big picture" issues without involving the Sixth Claimant. However, to obtain such order, the claimants would need to provide satisfactory evidence that future claims against the defendant will not arise, and they had not managed to do this by the hearing. Alternatively, the claimants could narrow the scope of Represented Parties for a more manageable set of legal determinations. The decision highlights a further difficulty with copyright claims relating to AI training where claimants need to find a way of putting together coherent large-scale representative actions.  It is perhaps not surprising that this issue has taken time to come into focus, given the range of issues that are involved in this "first of its kind" litigation. However, should others follow, there is a clear need for structured legal mechanisms to address mass copyright claims in AI training. It also underscores the benefit of clear consistent licensing agreements.  However, on the current state of the law, even if all the licenses had been identical, it still would not have addressed the very serious uncertainties surrounding the management of the case.

The parties must now resolve the case management issues quickly and prepare for the First Trial in June 2025.

The UK Government has recently launched a consultation on Copyright and AI (see our blog post here for link and commentary). This is considering requirements for transparency of AI training materials. Should transparency be enforced it would provide creators with specific evidence of use of their contents which could assist in providing a consistent basis for representative actions.