Shorter Reads

Getty Images’ claims against Stability AI are dismissed by UK High Court

In a closely watched case at the intersection of intellectual property and artificial intelligence, the UK High Court has dismissed Getty Images’ claims against Stability AI. The decision in Getty Images (US) Inc v Stability AI Ltd [2025] EWHC 38 (Ch), handed down on 4 November 2025, found that Stability AI did not infringe Getty’s copyright or trade mark rights in developing and operating its generative AI model, Stable Diffusion.

3 minute read

Published 7 November 2025

Authors

Share

Key information

In our previous article addressing how AI has impacted litigation, we considered the case of Getty Images (US) Inc v Stability AI Ltd [2025] EWHC 38 (Ch), and the implications it might have on the intersection of IP and AI, as well as the use of group actions.

The long-awaited UK High Court decision has now been handed down (on 4 November 2025) in which the claims of Getty Images were dismissed on various counts.

Background

By way of reminder, Getty is a visual media company who licenses stock photography images, and Stability is a developer of generative AI systems. This case concerns claims by Getty that Stability infringed its intellectual property (IP) rights by using substantial amounts of Getty’s content without permission to train and develop Stability’s AI models.

By the time the case reached trial, the scope of the claims against Stability had decreased significantly, leaving two key matters for determination.

Copyright Infringement

Stability AI had developed a deep learning model called Stable Diffusion. Getty’s copyright claim centred around whether the ‘AI model weights’ (the numerical parameters that determine how an AI model processes inputs to produce an output) for certain versions of Stable Diffusion amounted to an infringement of Getty’s copyright.

Initially, the copyright infringement claims could not be thoroughly assessed by the Court because Getty could not demonstrate that any of Stability AI’s training of its images had taken place in the UK. After Getty had to abandon those primary infringement claims, the thrust of Getty’s copyright claim was thus related to secondary infringement – in that even if the training was outside this jurisdiction, the importing of the infringing content into the UK made it illegal.

The Court also dismissed the secondary infringement claim. Whilst the AI model weights were found to fall within the definition of an ‘article’ (s22 and s23 Copyright Designs and Patents Act 1988 (CDPA)), Stable Diffusion also had to be proven to be an ‘infringing copy’ (s27 CDPA) which was imported into the UK. This meant the Court had to decide whether the model actually contains or embodies the copyright work, even momentarily. The AI model weights were not found to be infringing copies as they did not store or contain actual Getty images. The judge concluded that merely exposing model weights to infringing copies during training did not render the model itself an infringing copy.

Trade Mark Infringement

Getty Images had argued that Stability AI’s models violated its UK trade marks by producing synthetic images that included Getty or iStock watermarks. The court dismissed most of the trade mark infringement claims, noting that only three instances of infringement had been proved and there was no evidence of consumer confusion or commercial harm in such instances.

The court first examined whether there was evidence showing that UK users had generated synthetic images containing the watermarks. It concluded that there was enough evidence for older versions of Stable Diffusion (v1.x and v2.x), but not for newer models such as SD XL or v1.6. While the court therefore agreed that some trade mark infringement had occurred, the judge described these findings as “historic and extremely limited in scope”.

For the earlier versions, a small number of images with iStock watermarks were shown to have been generated in the UK. In these few cases, the court found infringement under section 10(1) Trade Marks Act 1994 (TMA) (use of identical marks for identical goods or services). Under section 10(2) TMA, which concerns use of identical or similar marks for identical or similar goods and services causing likely confusion, the judge identified limited instances, such as an image titled “Japanese Temple Garden”, where confusion could arise.

Finally, Getty’s claims under section 10(3) TMA, which deal with reputation-related harm, dilution, or unfair advantage, were rejected. The court found no convincing real-world evidence of reputational damage or change in consumer behaviour.

Comment

As one of the earliest IP claims against an AI developer to reach trial, this judgment inevitably offers guidance and clarity on the discrete issues which were addressed. But many practitioners are understandably disappointed that the judgement was nowhere near as comprehensive as anticipated when Getty’s claims were first launched. Following the extensive case management throughout, including 10 interim hearings, the issues which ultimately required determination at trial were narrowed significantly. This judgment could have been dramatically different, for example, had the alleged primary copyright infringement occurred within the UK.

But Stability will ostensibly take the win, as will other AI developers. Whilst the finding that an AI weighting can be an ‘article’ may have paved the way for a wider array of secondary infringement claims concerning other intangible methods of generating or storing copyright works, the judge’s strict analysis concerning whether the AI model constituted an ‘infringing claim’ demonstrates just how difficult that evidential burden will be to meet.

With many questions left unanswered we anticipate more litigation here in the UK, and certainly welcome further judicial decisions on this fast-evolving area.

Commentators are also linking the decision to the UK government’s latest consultation on the intersection of IP and AI, where they suggested the possibility of expanding the text and data mining exemption under the CDPA so that copyright protected works could be used for AI training, with an ‘opt out’ for rightsholders that would want to reject this mining. This proposal would widen the goalposts for what AI companies can do and further limit the rights of IP authors and owners. Finding a sufficient level of consensus for legislation to proceed in the UK has so far proved and insurmountable hurdle, but the Government may decide that it has to take a position, so we eagerly await further legislative developments.

Related latest updates
PREV NEXT

Related content

Arrow Back to Insights

Shorter Reads

Getty Images’ claims against Stability AI are dismissed by UK High Court

In a closely watched case at the intersection of intellectual property and artificial intelligence, the UK High Court has dismissed Getty Images’ claims against Stability AI. The decision in Getty Images (US) Inc v Stability AI Ltd [2025] EWHC 38 (Ch), handed down on 4 November 2025, found that Stability AI did not infringe Getty’s copyright or trade mark rights in developing and operating its generative AI model, Stable Diffusion.

Published 7 November 2025

Associated sectors / services

Authors

In our previous article addressing how AI has impacted litigation, we considered the case of Getty Images (US) Inc v Stability AI Ltd [2025] EWHC 38 (Ch), and the implications it might have on the intersection of IP and AI, as well as the use of group actions.

The long-awaited UK High Court decision has now been handed down (on 4 November 2025) in which the claims of Getty Images were dismissed on various counts.

Background

By way of reminder, Getty is a visual media company who licenses stock photography images, and Stability is a developer of generative AI systems. This case concerns claims by Getty that Stability infringed its intellectual property (IP) rights by using substantial amounts of Getty’s content without permission to train and develop Stability’s AI models.

By the time the case reached trial, the scope of the claims against Stability had decreased significantly, leaving two key matters for determination.

Copyright Infringement

Stability AI had developed a deep learning model called Stable Diffusion. Getty’s copyright claim centred around whether the ‘AI model weights’ (the numerical parameters that determine how an AI model processes inputs to produce an output) for certain versions of Stable Diffusion amounted to an infringement of Getty’s copyright.

Initially, the copyright infringement claims could not be thoroughly assessed by the Court because Getty could not demonstrate that any of Stability AI’s training of its images had taken place in the UK. After Getty had to abandon those primary infringement claims, the thrust of Getty’s copyright claim was thus related to secondary infringement – in that even if the training was outside this jurisdiction, the importing of the infringing content into the UK made it illegal.

The Court also dismissed the secondary infringement claim. Whilst the AI model weights were found to fall within the definition of an ‘article’ (s22 and s23 Copyright Designs and Patents Act 1988 (CDPA)), Stable Diffusion also had to be proven to be an ‘infringing copy’ (s27 CDPA) which was imported into the UK. This meant the Court had to decide whether the model actually contains or embodies the copyright work, even momentarily. The AI model weights were not found to be infringing copies as they did not store or contain actual Getty images. The judge concluded that merely exposing model weights to infringing copies during training did not render the model itself an infringing copy.

Trade Mark Infringement

Getty Images had argued that Stability AI’s models violated its UK trade marks by producing synthetic images that included Getty or iStock watermarks. The court dismissed most of the trade mark infringement claims, noting that only three instances of infringement had been proved and there was no evidence of consumer confusion or commercial harm in such instances.

The court first examined whether there was evidence showing that UK users had generated synthetic images containing the watermarks. It concluded that there was enough evidence for older versions of Stable Diffusion (v1.x and v2.x), but not for newer models such as SD XL or v1.6. While the court therefore agreed that some trade mark infringement had occurred, the judge described these findings as “historic and extremely limited in scope”.

For the earlier versions, a small number of images with iStock watermarks were shown to have been generated in the UK. In these few cases, the court found infringement under section 10(1) Trade Marks Act 1994 (TMA) (use of identical marks for identical goods or services). Under section 10(2) TMA, which concerns use of identical or similar marks for identical or similar goods and services causing likely confusion, the judge identified limited instances, such as an image titled “Japanese Temple Garden”, where confusion could arise.

Finally, Getty’s claims under section 10(3) TMA, which deal with reputation-related harm, dilution, or unfair advantage, were rejected. The court found no convincing real-world evidence of reputational damage or change in consumer behaviour.

Comment

As one of the earliest IP claims against an AI developer to reach trial, this judgment inevitably offers guidance and clarity on the discrete issues which were addressed. But many practitioners are understandably disappointed that the judgement was nowhere near as comprehensive as anticipated when Getty’s claims were first launched. Following the extensive case management throughout, including 10 interim hearings, the issues which ultimately required determination at trial were narrowed significantly. This judgment could have been dramatically different, for example, had the alleged primary copyright infringement occurred within the UK.

But Stability will ostensibly take the win, as will other AI developers. Whilst the finding that an AI weighting can be an ‘article’ may have paved the way for a wider array of secondary infringement claims concerning other intangible methods of generating or storing copyright works, the judge’s strict analysis concerning whether the AI model constituted an ‘infringing claim’ demonstrates just how difficult that evidential burden will be to meet.

With many questions left unanswered we anticipate more litigation here in the UK, and certainly welcome further judicial decisions on this fast-evolving area.

Commentators are also linking the decision to the UK government’s latest consultation on the intersection of IP and AI, where they suggested the possibility of expanding the text and data mining exemption under the CDPA so that copyright protected works could be used for AI training, with an ‘opt out’ for rightsholders that would want to reject this mining. This proposal would widen the goalposts for what AI companies can do and further limit the rights of IP authors and owners. Finding a sufficient level of consensus for legislation to proceed in the UK has so far proved and insurmountable hurdle, but the Government may decide that it has to take a position, so we eagerly await further legislative developments.

Associated sectors / services

Authors

Need some more information? Make an enquiry below.

    Subscribe

    Please add your details and your areas of interest below

    Specialist sectors:

    Legal services:

    Other information:

    Jurisdictions of interest to you (other than UK):



    Enjoy reading our articles? why not subscribe to notifications so you’ll never miss one?

    Subscribe to our articles

    Message us on WhatsApp (calling not available)

    Please note that Collyer Bristow provides this service during office hours for general information and enquiries only and that no legal or other professional advice will be provided over the WhatsApp platform. Please also note that if you choose to use this platform your personal data is likely to be processed outside the UK and EEA, including in the US. Appropriate legal or other professional opinion should be taken before taking or omitting to take any action in respect of any specific problem. Collyer Bristow LLP accepts no liability for any loss or damage which may arise from reliance on information provided. All information will be deleted immediately upon completion of a conversation.

    I accept Close

    Close
    Scroll up
    ExpandNeed some help?Toggle

    < Back to menu

    I have an issue and need your help

    Scroll to see our A-Z list of expertise

    Get in touch

    Get in touch using our form below.



      Business Close
      Private Wealth Close
      Hot Topics Close