Getty Images drops primary copyright claims against Stability AI

A preliminary judgement has been delivered by the High Court in an interesting case involving issues of copyright infringement, database right infringement, trade mark infringement and passing off. The Defendant company, Stability AI Ltd, offers a deep learning AI model for creating images generated by commands from users of its software.

The Claimant is the Getty Images Group of companies, a global visual content creator and licensor of images for customers across the world. Such content includes many original artistic works whose copyright is owned by or exclusively licensed to Getty Images.

Getty Images has claimed that Stability AI infringed its rights by downloading or ‘scraping’ images from its websites without Getty’s authorisation and inputting them into its AI model to ‘train’ it. Getty is also claiming that the images produced by the AI model also infringed its intellectual property rights since they contained a substantial part of Getty Images’ copyright protected works.

Various points were clarified by the Court in a recent hearing, including:

• recognition that training an AI model on copyrighted images, without a license to do so, can potentially infringe copyright, especially under UK law;
• recognition that UK copyright law may apply even if the data scraping or model training happens outside the UK but affects UK copyright holders.

This is a landmark case as it is one of the first to address AI training practices involving copyrighted data. The future rulings in this case will have significant implications for both AI companies and owners of copyright and online content, as they will deal with the issues of whether and how large-scale training models may breach copyright and other intellectual property rights.

In a key development on 25/06/25, Getty dropped its claim of primary copyright infringement against Stability AI at London’s High Court. This means Getty will no longer pursue the allegation that Stability directly copied its images in breach of UK copyright law.

Commentators suggest this may be due to difficulties in proving a sufficient connection between Stability’s actions and UK jurisdiction, as well as challenges demonstrating that the AI-generated images reproduced a ‘substantial part’ of Getty’s original works. While this narrows the scope of the dispute, secondary infringement and trade mark misuse claims, including allegations that AI-generated images unlawfully featured Getty’s watermark remain in play.

It remains to be seen what Stability AI’s liability is found to be and what remedies are afforded in the event of Getty Images’ case succeeding.

In the meantime, AI companies should take extra care when selecting data sources for training their AI models and consider whether existing sources present a risk of infringement of third party rights.

If possible, data should be taken from open-source or licensed content or from sources for which they have explicit permission to use. Due diligence should be conducted on all data sources and records kept of the provenance of all data employed. As companies may need to enter into additional licenses or decide to do so as best practice, they should factor the extra cost of this into their development budgets. AI companies should also consider purchasing legal insurance to cover potential litigation risks.

For business owners using AI or commissioning AI-generated content, this development highlights the importance of understanding how AI models are trained and where data originates. Although Getty’s primary infringement claim has been withdrawn, the remaining claims could still shape how intellectual property law applies to AI outputs in the UK. Businesses should be mindful of potential liability for unknowingly using infringing AI-generated materials and keep clear records when licensing or distributing such content.

The case may see a shift in bargaining power for copyright owners, which may entail strategic opportunities for negotiating licences and royalty payments. There is likely to be an increased demand for bulk licensing deals and for collections of data or works that are free to use or fully licensed.

Copyright owners should also do more to protect their works, ensuring that they are correctly archived and logged with ownership details and dates of creation and that copyright notices are clearly displayed and kept updated. There are many technologies that can be employed to track the unauthorised use of online content, such as digital watermarking and web monitoring tools. Auditing the potential exposure of copyright protected work to AI ‘scraping’ and use in datasets is recommended. Quick action should be taken against infringers, such as sending out cease and desist letters and takedown notices.