In a landmark ruling, the UK court has delivered a significant judgment regarding the intersection of artificial intelligence (AI) and intellectual property (IP) law, particularly focusing on the case between Getty Images and Stability AI. The court’s decision, rendered by Dame Joanna Smith, has sparked contrasting reactions, with some commentators deeming it a “damp squib,” while others recognize its pivotal implications for both the creative industries and generative AI developers.
Getty has hailed the ruling as a “significant win” for IP owners, citing the court’s findings of trademark infringement and affirming that the responsibility for unauthorized use of trademarks lies with the AI model provider rather than the end user. Stability AI, on the other hand, claims to have also emerged victorious, asserting that the judgment “ultimately resolves the copyright concerns” central to the case, having successfully defended the one remaining copyright claim. Looking ahead, Getty has set its sights on ongoing litigation in the United States, where it plans to leverage the findings from the UK case.
The court’s examination of the complex relationship between AI-generated synthetic images and trademark rights carries significant implications for the design of AI platforms. Notably, the ruling builds on previous cases that established platform liability for trademark use, such as the decisions involving Swatch. The judgment offers a nuanced exploration of copyright secondary infringement, addressing the challenges posed by the digital reproduction of creative works, as well as the issues surrounding model training and output liability.
Despite its immediate impact being somewhat understated, the ruling opens new avenues for potential disputes, particularly for claimants willing to invest significantly in combating unlicensed AI developers. As the UK government prepares to respond to its Copyright and AI consultation, the ruling underscores the ongoing tension between generative AI and rights holders.
The case originated in January 2023, when Getty Images initiated legal proceedings against Stability AI, alleging infringement of its intellectual property rights through the training and deployment of the generative AI model, Stable Diffusion. Getty contended that the datasets used to train the model included millions of images either owned or exclusively licensed to them.
Stability AI’s initial application for dismissal of the case was unsuccessful in October 2023, with three critical points at issue: jurisdiction regarding Stability AI’s actions in the UK, the classification of AI models under the Copyright, Designs and Patents Act 1988, and the use of watermarks in some synthetic images that allegedly infringed Getty’s trademarks.
The recent ruling provides an in-depth overview of how Stable Diffusion operates, delving into its training and output processes. The court recognized that the training of AI image generators relies heavily on paired images and descriptive metadata, making Getty’s extensive collection of images an invaluable resource for AI training. However, Getty significantly narrowed its claims during the trial, notably abandoning major copyright infringement claims associated with the model’s training and outputs.
While the court refrained from addressing whether the training of the model using copyrighted works constituted infringement—a point that Getty later conceded—other claims, particularly around secondary copyright infringement, were explored. This claim aimed to overcome territorial limitations on copyright infringement and suggested that unauthorized generative AI trained on copyrighted material could breach UK laws when such models are accessed.
Getty’s outputs claim, which sought to address the challenges posed by synthetic outputs that closely resemble original works, was similarly abandoned after Stability AI implemented measures to filter user inputs from generating outputs that might infringe copyrights. This indicates a tactical approach taken by Stability to limit potential litigation risks.
The trademark aspect of the case proved to be one of the more consequential parts of the judgment. The court found that users of Stability AI’s model could generate outputs featuring watermarks closely resembling Getty’s trademarks. The ruling determined that, as Stability AI did not effectively challenge the evidence of watermarked images generated, there was sufficient basis to consider trademark infringement under sections 10(1) and 10(2) of the Trade Marks Act. However, it found that certain versions of the model underwent adaptations that mitigated this risk.
Moreover, the court’s analysis regarding the context in which average consumers encounter these watermarks revealed that users are likely to pay attention to the outputs generated, which could influence their perception of the origin of the goods. This aspect could have broader implications for how generative AI companies approach their models and customer interfaces.
On the copyright front, the court’s conclusion regarding secondary infringement pivoted on the interpretation of “infringing copy” under the CDPA. The court held that the model weights, which are integral to the functioning of the model, did not constitute an infringing copy as they did not contain a copy of the copyrighted works despite the existence of prior infringement during the model’s training.
The ruling, while perhaps not delivering the comprehensive guidance desired by many in the AI and IP communities, nonetheless emphasizes critical considerations for AI developers in navigating the complexities of copyright law. As Getty prepares to take its findings from the UK case into its US litigation, the broader implications of this judgment will likely resonate throughout the creative and technological sectors.
As the UK government contemplates its policy response to the intersection of AI and copyright law, this ruling serves as a reminder of the urgent need to address the challenges and ambiguities presented by generative AI technologies in an evolving legal landscape.
See also
AI’s Environmental Impact Smaller Than Anticipated, May Propel Green Tech Advances
CoreWeave and AMD: Key Stocks Poised for Major AI Rallies by 2026
Top AI Video Editing Apps of 2025: Boost Quality and Efficiency for Creators
TSMC: The Top AI Stock to Buy Now with 30% Revenue Growth and Strategic U.S. Expansion
Genesis Mission Launches Ambitious AI Initiative to Transform Energy, Science, and Security
