Getty v Stability AI: First AI Judgment Made But Key Questions Remain Unanswered

In this blog post, authors Jonathan Ball & Polina Maloshchinskaia (Norton Rose Fulbright) look at the UK’s first major ruling on generative AI and intellectual property, Getty v Stability AI, decided on 4 November 2025. Much of the 205-page judgment focused on trade mark infringement, with the court finding that Stability AI could be liable where its models reproduced Getty’s proprietary watermarks in synthetic images. This confirmed that AI developers may bear responsibility for outputs that misuse protected marks.

The ruling’s more novel aspect concerned secondary copyright infringement. The judge held— for the first time since the 1988 Act—that an “article” under UK copyright law can include intangible items such as electronic copies. However, despite this expanded interpretation, the court rejected Getty’s secondary infringement claims, concluding that an AI model like Stable Diffusion does not store or reproduce actual copyright works and therefore cannot qualify as an “infringing copy.” Because the model itself is not a copy of any original work, Getty’s claims for importing, possessing, or dealing with infringing copies failed.

The judgment leaves major questions unresolved. Key issues around whether training a generative AI model on third-party works constitutes primary infringement, what remedies might apply if training occurred in the UK, and whether courts could impose injunctions or require model retraining remain unanswered, as those claims were not pursued at trial. The case also highlighted complications around copyright licensing, as Getty acted under New York–law licences that were not considered exclusive under English law, further limiting available remedies.

CONTINUE READING…