Two AI copyright cases, two very different outcomes

Home

Edition

Africa Australia Brasil Canada Canada (français) España Europe France Global Indonesia New Zealand United Kingdom United States

Edition: Global

Africa
Australia
Brasil
Canada
Canada (français)
España
Europe
France
Indonesia
New Zealand
United Kingdom
United States

s Newsletters

Academic rigour, journalistic flair

Feng Yu Two AI copyright cases, two very different outcomes – here’s why Published: December 1, 2025 2.00pm GMT Hayleigh Bosher, Brunel University of London

Author

Hayleigh Bosher
Reader in Intellectual Property Law, Brunel University of London

Disclosure statement

Hayleigh Bosher does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

Partners

Brunel University of London provides funding as a member of The Conversation UK.

View all partners

DOI

https://doi.org/10.64628/AB.wreajhqdf

https://theconversation.com/two-ai-copyright-cases-two-very-different-outcomes-heres-why-270229 https://theconversation.com/two-ai-copyright-cases-two-very-different-outcomes-heres-why-270229 Link copied Share article

Copy link Email Bluesky Facebook WhatsApp Messenger LinkedIn X (Twitter)

Print article

Artificial intelligence companies and the creative industries are locked in an ongoing battle, being played out in the courts. The thread that pulls all these lawsuits together is copyright.

There are now over 60 ongoing lawsuits in the US where creators and rightsholders are suing AI companies. Meanwhile, we have recently seen decisions in the first court cases from the UK and Germany – here’s what happened in those.

Getty Images, a global visual content creator and marketplace, sued Stability AI, an open-source generative AI company, in the UK courts. Getty claimed Stability had illegally used its content to train an AI model called Stable Diffusion. Getty is also suing Stability AI in the US and that case is on-going.

It was accepted that Getty’s images were used in the training of Stable Diffusion without permission, and that this training involved copying. Copyright is the right to stop someone copying your work, so training an AI model on copyright-protected content without permission would be an infringement of copyright under UK law.

But the plot twist here is that the training of Stable Diffusion took place outside the UK, meaning that Getty ended up narrowing their copyright case to focus on what is called “secondary infringement” – which is essentially the same as importing goods that infringe copyright. It is illegal to bring counterfeit DVDs into the UK for the same reason.

At trial, the judge had to decide something for the first time ever. This was: if a user downloads an AI model in the UK that was illegally trained on copyrighted content in another country, does that count as secondary copyright infringement? To do so, she had to consider two things. The first was: can the definition of an “article” include intangible goods?

When someone imports a box of counterfeit DVDs, these are called “articles” under the law, and are obviously physical, tangible objects. Stability AI argued that its AI model (system) was not an “article” because it is not a physical object. The judge, sensibly, understood that the law was written long before the new era of AI, but the intention behind the rule was to include both tangible and intangible goods.

The second thing the judge had to consider was: is the Stable Diffusion AI model what copyright law calls an “infringing copy”?

The judge took this to mean the model would need to physically contain a reproduction of Getty’s content. But the way the model “learned” from the training data, according to the expert evidence, meant that it did not actually contain any copies. So, Getty lost the claim for secondary copyright infringement.

Stable Diffusion is an AI art generator developed by Stability AI. Olivia_Brown

Differing interpretations

Meanwhile, one of the world’s largest collecting societies for musical works, GEMA, filed a copyright lawsuit against the company OpenAI in Germany. This was for the use of song lyrics in its large language model ChatGPT.

On November 11 2025, the Munich court decided, like the UK court, that training AI on copyrighted content requires a licence. However, it took a different approach to interpreting the law of copying, and essentially said that since the AI model was trained on the lyrics and could reproduce them as an output, then the model had embodied the content.

Unlike the UK decision, the German court found the technical way in which the AI model does this to be irrelevant, so GEMA won the case.

One thing both courts did agree on is that the AI developers are liable for any infringement, not – as the AI developers argued – the users who select a prompt which the model then responds to by generating content.

Although the circumstances of the cases are slightly different, the heart of the issue was the same. Do AI models reproduce copyright-protected content in their training process and in generating outputs? The German court decided they do, whereas the UK court took a different view.

Both cases could be appealed and others are underway, so things may change. But the ending we want to see is one where AI and the creative industries come together in agreement. This would preferably happen with the use of copyright licences that benefit them both.

Importantly, it would also come with the consent of – and fair payment to – creators of the content that makes both their industries go round.