The High Court delivered its judgment in Getty Images v Stability AI on November 4, 2025, marking a pivotal moment in the UK’s legal landscape concerning generative AI and intellectual property law. The case, seen as a landmark, narrowed in scope as Getty Images dropped several claims; however, it still addressed critical issues regarding the application of intellectual property rights in the era of generative AI, particularly focusing on trade mark matters.
In the ruling, the court dismissed Getty’s assertion that Stability AI’s UK activities constituted secondary copyright infringement. The training of the Stable Diffusion model occurred outside the UK, and its subsequent importation and distribution did not count as dealing in infringing copies. Nevertheless, on December 16, 2025, Mrs. Justice Joanna Smith granted Getty permission to appeal this aspect to the Court of Appeal.
On the other hand, Getty achieved a limited victory regarding trade mark infringement. The court recognized some historic instances of trade mark infringement under sections 10(1) and 10(2) of the Trade Marks Act 1994, primarily related to watermarks appearing in specific outputs. However, it rejected claims under section 10(3) and chose not to rule on passing off cases. Following the judgment, Mrs. Justice Joanna Smith denied Stability AI’s request to appeal her findings related to trade mark infringement.
Getty Images employs watermarks in its images featuring the phrases “GETTY IMAGES” and “iSTOCK,” both of which are registered trade marks. The Trade Marks Act outlines three scenarios for trade mark infringement, evaluated through the lens of the ‘average consumer’ or ‘relevant public.’ Section 10(1) prohibits the use of a sign identical to a registered trade mark concerning identical goods or services, while section 10(2) disallows the use of a sign similar to a trade mark with the likelihood of confusion concerning similar goods or services. Section 10(3) protects the trade mark owner’s reputation, covering cases where the goods and services differ but the use of the sign could lead to unfair advantage or harm the mark’s reputation.
Getty’s claim against Stability AI centered on the appearance of watermarks, which it argued were either identical or similar to its registered marks, present in the synthetic image outputs produced by the Stable Diffusion model. Although not every output contained such watermarks, Getty asserted that it had evidence showing at least some outputs did. To substantiate its claims, Getty needed to establish that at least one infringing output was generated in the UK. The court conducted a detailed, model-specific examination and found “historic” and “extremely limited” instances of infringement under sections 10(1) and 10(2), while dismissing claims under section 10(3).
The court identified three classes of average consumers: those using the downloadable mode, those engaged with the developer model, and those accessing the web-based model, depending on their technical proficiency. The assessment of infringement was based on the version-by-version analysis of Stable Diffusion. This granular method aligned with the technical nature of Stability’s models, which were trained on varying datasets with different filters. Evidence indicated that later iterations of Stable Diffusion incorporated filters designed to exclude watermarked images from its training dataset, resulting in a reduced likelihood of generating outputs with watermarks.
Although Getty’s experimental prompts designed to trigger watermarked outputs faced scrutiny, the court acknowledged that some generic prompts, like “news photo” and “vector art,” realistically reflected user behavior. Getty also provided instances of outputs generated using random prompts; while the UK origin of these examples couldn’t be definitively proven, the court considered them indicative of likely UK generation.
Regarding section 10(1), the court concluded that the provision of synthetic image outputs fell under class 41 for “digital imaging services” and class 9 for “downloadable digital illustrations and graphics,” affirming infringement. However, the court dismissed Getty’s argument that these outputs qualified under “computerized online search and retrievable services for images” in class 42.
In evaluating section 10(2), the court considered the implications of post-sale confusion, in line with Supreme Court guidance from Iconix v Dream Pairs. The court determined that evidence pointed toward a significant proportion of the relevant public experiencing confusion, noting that many consumers associate watermarks with paid content. As a result, the court found limited outputs infringed under section 10(2).
The judgment emphasized that Stability AI’s actions constituted “use in the course of trade.” While users could modify prompts, they had no control over watermark appearances, which stemmed from Stability’s design choices rather than user intent. Conversely, Getty’s arguments involving section 10(3) related to dilution, tarnishment, and free-riding were dismissed, as the court found no evidence of detrimental impact on the mark’s distinctive character or reputation, nor any unfair advantage.
This ruling clarifies that trade mark infringement in the context of generative AI is heavily fact-dependent, particularly concerning the nature of user prompts and control. As brands increasingly pursue trade mark infringement claims against AI developers, the authenticity of user prompts and the variability of allegedly infringing outputs will likely play crucial roles in future assessments. In this evolving landscape, the authenticity of user intent and the consistency of outputs may become vital elements in determining the viability of such claims.
See also
Germany”s National Team Prepares for World Cup Qualifiers with Disco Atmosphere
95% of AI Projects Fail in Companies According to MIT
AI in Food & Beverages Market to Surge from $11.08B to $263.80B by 2032
Satya Nadella Supports OpenAI’s $100B Revenue Goal, Highlights AI Funding Needs
Wall Street Recovers from Early Loss as Nvidia Surges 1.8% Amid Market Volatility
