A recent court ruling in London has added another chapter to the rapidly evolving legal landscape surrounding artificial intelligence and copyright law, as creative industries and AI companies grapple with questions about the use of human-produced content to train powerful machine learning models.
The case at the center of this attention was brought by Getty Images, one of the world’s most prominent stock photo agencies, against Stability AI, the company behind the widely used Stable Diffusion image generation model. Getty Images alleged that Stability AI unlawfully used its copyrighted content to train their AI models, essentially “scraping” images from Getty’s vast online database without permission or compensation.
On Tuesday, Justice Joanna Smith delivered her decision and, in doing so, highlighted the complexity and nuance that often define AI-related copyright cases. Her ruling found that Stability AI did not infringe upon Getty’s copyrights because the company did not “store or reproduce any Copyright Works and nor has it ever done so.” In other words, the process of training an AI model on images—at least as Stability AI had done—did not in itself constitute a copyright violation under UK law.
However, Justice Smith was careful to emphasize that her ruling was both “historic” and “extremely limited” in its scope. Instead of issuing a sweeping decision that would set broad guidelines for the entire AI industry, she focused narrowly on the facts and legal arguments presented in this particular case. This approach mirrors recent rulings in US courts, where judges have also been cautious about creating wide-reaching precedents in the fast-moving area of AI and copyright.
Despite Stability AI’s win on the core copyright issue, the court did find in Getty’s favor on a separate point related to trademark law. Getty had argued that Stable Diffusion, when generating new images, sometimes included elements that closely resembled the iStock and Getty Images logos. Justice Smith agreed that allowing users to produce AI-generated images with these trademarks could amount to trademark infringement under certain legal statutes. Importantly, she clarified that the responsibility for this kind of infringement lies with the model provider—Stability AI—and not with the individual users generating the images. This point reinforces the idea that companies building and distributing AI tools bear a significant responsibility for how their models are trained and the outputs they produce.
Both sides of the dispute found aspects of the ruling to celebrate. Getty Images hailed the decision as a win for intellectual property holders, emphasizing the court’s recognition of the risks posed by AI-generated images that mimic protected trademarks. In their statement, Getty praised the court’s determination that AI companies must be held accountable for the content their models are able to generate, rather than shifting liability onto users.
Stability AI, for its part, pointed out that the majority of Getty’s original copyright claims had been dropped before the final stage of the trial, leaving only a subset of issues for the court to resolve. Christian Dowell, Stability AI’s general counsel, stressed that the final ruling addressed and resolved the copyright concerns that had initially been at the heart of the lawsuit