Trial Outcome between Getty Images and Stability AI: Copyright Implications Explored
In a groundbreaking legal battle, Getty Images is taking on Stability AI in a case that could shape the future of copyright law and AI technology. The case, known as Getty v. Stability AI, has shed light on the complexities of applying traditional copyright laws to AI-generated content.
The battle began during the training and development stage, where Getty initially claimed copyright infringement against Stability AI. However, due to a lack of evidence and jurisdictional challenges, Getty dropped its primary claims. Most of Stable Diffusion's training occurred on servers in the United States, outside the UK's jurisdiction, making it difficult for Getty to prove infringement within the UK.
The case also highlighted the difficulty in determining where AI training occurs, which is crucial for establishing jurisdiction, and the evidentiary challenges in proving copyright infringement, especially when the training process is complex and distributed across multiple locations.
Moving on to the output stage, Getty dropped its claims regarding the AI-generated images, as Stability AI implemented measures to prevent the generation of infringing content. The case underscores the importance of technical solutions in mitigating liability and the challenge of proving that AI-generated images reproduce a substantial part of protected works, which is essential for establishing copyright infringement.
The importation stage was not the primary focus of the case, but the jurisdictional and evidentiary challenges faced by Getty could similarly apply to any stage involving cross-border activities.
Getty continues to pursue other claims, including whether Stable Diffusion itself constitutes an infringing article under copyright law, and trademark infringement claims related to the appearance of Getty watermarks in AI-generated images. The case will influence the development of AI law, particularly in how copyright legislation is interpreted and applied to AI models.
The complexities of the wilful contrivance issue were extensively discussed, but no ruling was made. The remaining claim is of secondary infringement, concerning the importation, possession, and/or distribution of an infringing copy. The difficulty for Getty arises from the nature of latent diffusion models, which do not always produce the same image with the same prompt.
Getty alleged that datasets containing infringing copies of its works were downloaded and stored by Stability in the UK during the training and development of Stability's model. However, the ruling on secondary infringement will be highly significant, potentially impacting content and services provided to UK consumers from overseas.
Establishing that a substantial part of a protected element has been taken is challenging due to the low originality of some images and the lack of a clear memorized output. The judge might have to address this question as part of the trademark infringement claim.
The judge found "strong support for a finding that, on the balance of probabilities, no training or development" took place in the UK, but also held that there was evidence potentially pointing away from this. The primary infringement by copying during the training and development stage was dropped. A finding of liability could be limited to downloadable versions of the model, leaving Stability free to supply web-based versions hosted outside of the UK.
The ruling on secondary infringement is unlikely before the summer recess but is expected to be comprehensive. The case underscores the complexities of applying traditional copyright laws to AI technologies, highlighting the need for clearer legal frameworks to address these emerging issues.
- Getty's allegation of using datasets containing infringing copies of its works during Stability AI's training phase raises questions about copyright compliance, particularly in the context of AI technology and the importation, possession, and distribution of infringing copies.
- The complexities surrounding the application of traditional copyright laws to AI-generated content were evident in the discussion of whether Stable Diffusion constitutes an infringing article under copyright law, with the issue of originality and memorized output being a key point of consideration.
- The judge's ruling on secondary infringement, potentially impacting content and services provided to UK consumers from overseas, will be significant, emphasizing the need for legal clarity in the rapidly evolving field of AI technology and its relationship with copyright law and patents.
