On 16 December 2025, Mrs Justice Joanna Smith DBE granted permission for Getty Images to appeal her judgment in November 2024, which dismissed Getty’s copyright infringement claims against Stability AI. The opportunity for the Court of Appeal to guide the world and the UK on the correct understanding of copyright and AI is very significant. The appeal will be closely watched by all of those concerned with copyright and AI.
Getty Images launched proceedings in the High Court in 2023 against Stability AI, originally alleging infringement of copyright in millions of images used to train Stability’s Stable Diffusion AI, as well as infringement of database rights and trade marks. The case pushed into uncharted territory on the use of copyright materials in the training of AI models.
As the proceedings progressed, Getty narrowed the scope of its claims. Prior to judgment in November 2024, Getty decided not to pursue its claim of primary infringement, the central challenge being that training had taken place entirely outside the UK and thus principally outside the reach of UK copyright law under the Copyright Designs and Patents Act 1988 (CDPA).
Getty instead rested its copyright claim on whether the offering of access to Stable Diffusion in the UK constituted the importation of an infringing article by Perplexity. The High Court judgment held that whilst an ‘infringing article’ may consist of intangible property, the model weights underpinning Stable Diffusion did not reproduce Getty’s images, and so did not constitute an infringing article.
Permission to appeal
In granting permission to appeal her decision, Mrs Justice Joanna Smith DBE acknowledged that the Court of Appeal may arrive at a different conclusion on the question of secondary infringement and that Getty’s proposed appeal “does in [her] judgement, have a real prospect of success. It concerns a pure question of law, namely a matter of statutory construction on which the minds of reasonable lawyers may differ”.
The judge also recognised the broader importance of the answer to the questions at the heart of the dispute:
“… there are other compelling reasons for the appeal. The point of law is both novel and important because it concerns how the provisions of the CDPA should be construed (and specifically the phrase “infringing copy”) in the context of an AI model. This is not an issue that has previously been considered by any court. As the claimants submit, that novel point also has potentially far-reaching ramifications for AI models and intangible articles such as software more generally.”
Importantly, the judge also ensured that the door is left open for the UK courts to consider in future whether the use of copyright images to train an AI constitutes an act of primary infringement, by requiring language in the final order to the effect that whilst Getty’s claims were abandoned late in the proceedings – indeed after considerable evidence and cross-examination – they were not adjudicated upon by the court.
The broader importance of the legal position on copyright and AI training is underscored by the level of public and industry response to the UK government’s consultation on possible reforms to copyright law. In December 2025, the government published the results of its consultation, which drew over 11,500 responses. An overwhelming majority of 88% favoured the need for a licence to use copyright protected works for training AI.
What is happening elsewhere on copyright and AI?
In early November, the Munich Regional Court issued a ruling in favour of the German musical royalty collecting society, GEMA, holding that OpenAI had infringed the copyright of ruled that OpenAI had infringed copyright in the lyrics to a song which were said to have been used to train its ChatGPT models. The judgement sets the EU on a diverging path to the UK on the use of copyright works for training AI models.
GEMA argued that the lyrics had been memorised by the ChatGPT models, allowing them to be reproduced nearly identically in response to relatively simple prompts. The Court found that this satisfied the requirement for fixation under EU copyright law. It found, therefore, that the storage of the lyrics in model parameters, and their subsequent reproduction in outputs, constituted infringements of copyright.
In addition to the appeal which OpenAI plans to pursue, a reference to the Court of Justice of the European Union is a possibility, including on the scope of the EU’s text and data mining (TDM) exception under the EU’s Directive on Copyright in the Digital Single Market.
In the US, a class proceeding on behalf of authors in Bartz v Anthropic held a similar promise of being a watershed moment for copyright and AI. The plaintiff authors alleged that Anthropic had infringed copyright in works that were used to train the company’s Claude AI models. Those works were obtained by Anthropic in two distinct scenarios, lawfully, on the one hand, and via libraries of pirated materials, on the other.
US copyright law is typically more permissive on the use copyright works for new purposes, relying on a relatively doctrine of fair use, as opposed to the UK’s more rigid approach to fair dealing. In June 2025, a judge held that Anthropic’s use of books obtained lawfully for AI training was ‘transformative’ and constituted fair use under US copyright law. The use of pirated copies of books was held not to be transformative, and so the plaintiffs were able to maintain their claims against Anthropic in that regard.
The US statutory copyright regime meant that Anthropic faced potentially existential damages in the billions of dollars if the authors’ claim succeeded. In September 2025, a settlement was announced whereby Anthropic would pay at least $1.5 billion dollars, equating to roughly $3,000 for each of the pirated book copies at issue, setting a first potential benchmark for royalty levels in licensing arrangements.
If you have questions or concerns about the topics raised in this article, please contact James Tumbridge and Robert Peake.