Getty’s Copyright Case Against Stability AI Fails
The High Court in London has ruled that content creators and publishers can only succeed in claims of secondary copyright infringement against AI developers in the UK if the AI systems trained on their content actually store or reproduce their works.
AI Copyright Decision
Getty Images claimed that Stability AI used millions of its photographs to train its image-generation model, and attempted to claim secondary copyright infringement under the Copyright, Designs and Patents Act 1988 by alleging that the AI model qualified as an "article" imported into the UK that was an infringing copy.
In the case, Getty alleged copyright infringement, database right infringement, trade mark infringement and passing off. During the trial, Getty withdrew its primary copyright infringement (unauthorised copying or communication to the public) and database right infringement claims because of evidential and jurisdictional weaknesses.
While the claim of secondary copyright infringement (importing, marketing, selling and distributing) remained, it was rejected by the Court because the model (Stable Diffusion) did not store or reproduce Getty's photographs.
However, a partial victory for Getty did arise in relation to trade mark infringement for some watermarks from early versions of the model.
Implications for Content Creators and Publishers
For content creators and publishers, this ruling means that claims based solely on the fact that works were used in training datasets will face steep obstacles in the UK without evidence showing that the AI system stored or reproduced substantial part of the claimant's works. Where training and storage occurred outside the UK, jurisdictional and evidential hurdles may apply as demonstrated in this case.
Secondary copyright infringement may remain viable, but only in circumstances where the AI model or system can be shown to contain or reproduce the works in question.
Accordingly, content owners should reassess their rights-management strategies, auditing how their works may be used in AI training and exploring collective licensing or sector-wide approaches.
Implications for AI Developers
For AI developers, this ruling offers a degree of comfort that an AI model transforming data during training (without storing or reproducing the underlying copyright works) is not automatically liable for secondary copyright infringement under UK law.
However, caution remains advisable by keeping transparent records of data-sourcing, training processes and transformations. As this ruling leaves open significant questions, especially around primary copyright infringement and jurisdiction, further legal developments should be monitored closely and AI developers should be alert to other IP risks (such as trade mark infringement) as evidenced in the Getty case.
Tidman Comment
Since a consultation on possible changes to UK copyright law for the AI age last winter, the UK government has established working groups to help identify practical solutions on matters of transparency and control of rights. However, it is not expected to issue a substantive progress update until around Christmas.
In the meantime, this court decision is significant for its delineation of the threshold for secondary infringement in the AI-training context.
Make an Enquiry Now
We offer confidential, practical advice tailored to your business. Call our expert copyright experts on 0131 478 4724 or complete an Online Enquiry.
We have helped hundreds of individuals and businesses identify and protect their rights over product creation in the UK.
Please note the contents of this blog is given for information only and must not be relied upon. Legal advice should always be sought in relation to your specific circumstances.
