Background of the Case
The legal dispute between Getty Images and Stability AI marked the first major UK judgment addressing copyright infringement in the context of AI training. The case centered on whether Stability AI’s image-generating model, Stable Diffusion, infringed Getty Images’ copyright by using its photographs as part of its training dataset. The ruling, delivered on November 4, 2025, clarified that the AI-generated outputs themselves were not direct copies of Getty’s works, but it did not definitively resolve the broader question of whether the training process itself constitutes copyright infringement. This issue remains pending for legislative clarification by Parliament.
The case was heard in the High Court of England and Wales, with the judgment setting a precedent for future disputes involving AI and copyright.
Getty Images International U.C., a New York-incorporated company, represented the plaintiff. The legal proceedings focused on whether the AI model’s training process—specifically the use of copyrighted images—constituted infringement under UK copyright law. The court’s decision emphasized that the outputs generated by the model were not direct copies, but it left unresolved whether the training process itself required legislative intervention to define its legality.
Key Legal Arguments and Findings
Getty Images’ central argument was that Stability AI’s use of its photographs in training violated Section 16 of the UK Copyright, Designs and Patents Act 1988, which prohibits unauthorized copying of copyrighted works.
The court ruled that the AI-generated outputs did not constitute direct copies of Getty’s photographs. Instead, the judgment noted that the model’s outputs were novel creations, distinct from the original works used in training. However, the court did not address whether the act of training the AI model on copyrighted material—without explicit permission—was lawful. This omission highlighted the need for legislative reform to clarify the boundaries of AI training practices under copyright law.
Implications for AI Development and Copyright Law
While the court did not find infringement in the outputs, the unresolved question of training legality creates uncertainty for AI developers.
The judgment serves as a reminder that current copyright frameworks were not designed with AI in mind. The rapid advancement of AI technologies has outpaced existing legal structures, leaving gaps that require legislative attention. Parliament’s pending reform of copyright law is critical to providing clear guidelines for AI training practices, ensuring that innovation can proceed without the threat of litigation.
While the court did not rule in favor of Getty on the specific issue of training, the judgment may encourage other copyright holders to pursue similar claims, particularly if legislative reforms do not adequately address their concerns.
Conversely, if no clear rules are established, copyright holders may face continued unauthorized use of their works.
Pending Legislative Reform and Future Outlook
Stability AI judgment did not provide a definitive answer on the legality of AI training, it has accelerated discussions around how to update copyright frameworks to accommodate AI technologies. Parliament is expected to introduce legislation that clarifies the boundaries of AI training, potentially including provisions for fair use, licensing requirements, or other exceptions tailored to AI development.
This could allow copyright holders to control how their works are used in AI training while providing AI developers with clearer legal parameters. Alternatively, the reform may introduce a new exception for AI training, provided that certain conditions—such as compensation to copyright holders—are met.
For AI companies, the pending reforms present both risks and opportunities. On one hand, uncertainty over training legality could lead to costly litigation or operational disruptions.
The judgment has demonstrated that legal action can be an effective tool for challenging unauthorized AI training, but long-term solutions will require proactive participation in policy discussions.
Required actions
- An AI model that does not store or reproduce the works it was trained on is not an "infringing copy" under the CDPA, so importing or distributing it is not secondary infringement - Training on copyrighted works outside the UK is not actionable under UK law; watermark reproduction can raise trade mark issues