- with Senior Company Executives, HR and Finance and Tax Executives
- in Canada
- with readers working within the Banking & Credit, Business & Consumer Services and Healthcare industries
Key Takeaways
- The U.K. High Court ruled that Stable Diffusion is not infringing despite being trained on potentially copyrighted works.
- The decision may influence AI copyright cases in Canada due to similarities in copyright laws between the U.K. and Canada.
- The case highlights challenges in class actions for copyright claims, requiring individualized assessments for each work involved.
The U.K. High Court of Justice (the High Court) issued an important decision on November 4 that will be well-received by developers, distributors and users of AI models and systems and a disappointment to the creators and owners of artistic works.1
Most significantly, the Court found that Stable Diffusion, a generative AI tool that creates high-quality images from text descriptions, is not an infringing article merely because it was trained using a process that Getty Images claimed involved copyright infringement. Rather, the Court concluded that the model weights comprising a model never store or reproduce the copyrighted works used in training but are instead a product of the pattern and features learned over time during the training process.
The case also demonstrates some of the challenges associated with identifying output from a generative AI model that infringes on the rights in works used to train it. The High Court held in a January 2025 ruling2 that proving a text prompt claim requires a case-by-case comparison of each allegedly infringing output to a specific work of the plaintiff to determine whether the former reproduced a substantial part of the latter. Getty Images ultimately abandoned its claim that outputs generated using Stable Diffusion were infringing copies of its works.
Getty Images did have some success in making out trademark infringement claims. It alleged that early versions of Stable Diffusion generated outputs containing Getty and iStock watermarks. The Court rejected some of these claims due to lack of evidence that users in the U.K. ever generated the watermarks in question. While other claims succeeded, the Court characterized them as "both historic and extremely limited in scope."
The High Court's decision did not address the validity of Stability AI's argument that the development and training of Stable Diffusion outside the U.K. does not implicate U.K. copyright law. This issue became moot when Getty Images withdrew its allegations of infringement in respect of model training, following an unfavourable ruling on an amendment Getty Images sought to make to its pleading shortly before trial.3
Implications for AI developers and content owners in Canada
The High Court's decision is expected to be closely read beyond the U.K. While copyright is territorial and U.K. judgments are not binding in other jurisdictions, the case is one of the first generative AI copyright disputes to reach trial. Similarities between Canadian and U.K. copyright laws will also make the decision of particular interest in Canada, including in connection with a growing number of AI copyright cases in this country.
The Stable Diffusion case also highlights substantive and procedural challenges that arise in class proceedings involving allegations of copyright infringement. By way of example, in a pre-trial ruling, the High Court rejected a "representative" claim by Getty Images, finding that the assessment of whether any licensor fell within the proposed class required an individualized assessment of whether their works had been used to train Stable Diffusion. The U.K. court concluded that this made the definition of a class impossible.
Footnotes
1. Full decision available at: https://www.judiciary.uk/judgments/getty-images-v-stability-ai/
2. Pre-trial decision available at: https://www.judiciary.uk/wp-content/uploads/2025/11/Getty-Images-v-Stability-AI.pdf
3. Pre-trial decision available at: https://www.judiciary.uk/wp-content/uploads/2025/11/Getty-Images-v-Stability-AI.pdf
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
