UK Landmark Decision: No Infringement Without Embodiment
In a groundbreaking ruling, the UK High Court delivered its first major judgment on generative AI and copyright in Getty Images (US) Inc & ors v Stability AI Limited ([2025] EWHC 2863 (Ch)) [6]. The court held that Stability AI did not infringe Getty Images’ copyright through its training of the Stable Diffusion model, because the resulting model weights, while derived from copyrighted images, do not contain or embody those images in any tangible or transient form [1].
The court emphasized that for an article to be an infringing copy under UK law, it must actually contain the protected work [5]. Since the model weights consist of numerical parameters shaped by statistical learning rather than stored reproductions, they cannot constitute “infringing copies” under Section 17 of the Copyright, Designs and Patents Act 1988 [3]. As one analysis noted: “The Stable Diffusion models do not contain or store reproductions of the relevant works on which they were trained” [2].
Notably, the court found that Stability AI likely infringed copyright during the training phase if those acts occurred in the UK, but the actual model’s importation and distribution were not unlawful because the final product did not embed the original works [31]. This sharp distinction between process and product forms the crux of the UK’s approach: infringement hinges on whether the output, or the model itself, embodies the work [9].
A detailed memorandum from Sidley Austin called the decision a clear win for AI developers, reinforcing that use of copyrighted material in training, absent retention in the final model, is not secondary infringement under UK law [29].
US Landscape: Litigation in Full Swing
In contrast, US courts have yet to issue a final ruling, and the legal landscape remains unsettled. High-stakes cases are actively moving forward, with no clear resolution in sight.
The New York Times v. OpenAI and Microsoft has emerged as a pivotal case, with the Times alleging that the companies used millions of its articles to train GPT models without permission, constituting copyright infringement [12]. A federal judge has allowed the lawsuit to proceed, rejecting OpenAI’s bid to dismiss the claims [16]. The case is widely viewed as one of the first major tests of fair use and reproduction rights in the AI era [14], with OpenAI defending its actions under transformative fair use while resisting discovery demands that would expose private user data [10].
Another critical case is Andersen v. Stability AI Ltd., filed in January 2023 by artists Sarah Andersen, Kelly McKernan, and Karla Ortiz, and supported by members of DeviantArt and other platforms [19]. This class-action lawsuit challenges the use of artists’ works to train Stable Diffusion without consent, alleging direct copyright infringement, violation of the DMCA, and right of publicity claims [20]. A California judge has allowed the case to move forward, signaling skepticism toward AI firms’ broad claims of lawful use [22]. Notably, this U.S. litigation directly mirrors the UK dispute but proceeds under a different legal framework that gives more weight to whether training itself, regardless of output, constitutes reproduction [26].
Anthropic’s $1.5 billion settlement in a separate US case, though not publicly detailed, reflects the high financial stakes and risk-aversion among AI firms facing potential liability under US law [citation:N/A].
Divergent Paths: One Issue, Two Standards
The UK and US are now on divergent legal trajectories. The UK High Court’s narrow, embodiment-based interpretation provides AI developers clarity: if the model doesn’t contain or transmit the original work, it’s not an infringing copy [4]. This safe harbor for statistical abstraction may encourage innovation within the UK and influence other common law jurisdictions.
In contrast, US courts are still grappling with foundational questions: Does copying during training violate the reproduction right? Is such use “fair” under Section 107? And can outputs that resemble copyrighted works trigger liability even if the input isn’t stored?
With ongoing cases across media [press (NY Times), visual art (Andersen/DeviantArt), and software] US jurisprudence could evolve toward broader liability or robust fair use protections. Until then, AI firms operate under greater uncertainty in the US than in the UK [36].
Outlook and Implications
The split between the UK and US approaches highlights the global fragmentation of AI copyright law. Firms developing generative AI must now navigate a patchwork of national standards: training may be permissible in London but legally perilous in San Francisco.
Policymakers on both sides of the Atlantic are watching closely. The UK’s judgment may spur calls for legislative updates to close perceived loopholes [36], while the US Copyright Office continues its own review of AI-related policy, offering non-binding guidance in the interim [5]([35]).
As the legal battlelines solidify, one truth emerges: the future of AI may not be shaped by code—but by courts.
2.Getty Images v Stability AI English High Court Rejects …
3.Getty Images v Stability AI: What the High Court’s Decision …
4.Getty Images v. Stability AI: Intellectual Property Rights in …
5.Pixels at Dawn: Getty Images v Stability AI
6.First Major Judgment on How UK Copyright Law Applies to …
7.Getty Images v Stability AI: The UK Courts’ First Word on …
8.Getty Images v Stability AI: the copyright ruling
9.Getty Images (US) Inc (and others) v Stability AI Limited. Input
10.How we’re responding to The New York Times’ data …
11.From Copyright Case to AI Data Crisis: How The New York …
12.NYT v. OpenAI: The Times’s About-Face
13.Lawsuits Blame ChatGPT for Suicides and Harmful …
14.Does ChatGPT violate New York Times’ copyrights?
15.OpenAI slams court order that lets NYT read 20 million …
16.Judge allows ‘New York Times’ copyright case against …
17.Stolen Stories or Fair Use? The New York Times v. OpenAI …
18.New York Times v. Microsoft
19.Andersen v. Stability AI: The Landmark Case Unpacking …
21.Andersen v. Stability AI Ltd.
22.Artists Land a Win in Class Action Lawsuit Against A.I. …
23.Copyright in the Age of Generative AI, Part II
24.Andersen v. Stability AI Ltd., — F.Supp.3d —- (2024) – CDN
25.Class Action Lawsuit v. Stability AI, DeviantArt, Midjourney …
26.Andersen et al v. Stability AI Ltd. et al, No. 3:2023cv00201
27.Takeaways from the Andersen v. Stability AI Copyright Case
28.The UK’s First Copyright vs. AI Decision
29.Sidley Discusses the UK’s First Copyright vs. AI Decision
30.Sidley Discusses the UK’s First Copyright vs. AI Decision
31.UK High Court rules AI creation not a copy of original material
32.Getty Image Loses Copyright Infringement Claim Against …
33.Getty Images v Stability AI English High Court Rejects …
34.UK High Court: landmark decision on copyright and Gen AI
35.Getty Images issues statement on ruling in Stability AI UK …