In another Big AI Battle, this time in the UK, Getty Images brought an action against Stability AI. Getty is the owner (or exclusive licensor) of many photos, and Stability AI is the owner of the text-image AI model known as Diffus. The case is probably the most significant AI case so far albeit confined.
The key policy debate is neatly summarised by MRS JUSTICE JOANNA SMITH at paragraph 12:
“Both sides emphasise the significance of this case to the different industries they represent: the creative industry on one side and the AI industry and innovators on the other. Where the balance should be struck between the interests of these opposing factions is of very real societal importance. Getty Images deny that their claim represents a threat to the AI industry or an attempt to curtail the development and use of AI models such as Stable Diffusion. However, their case remains that if creative industries are exploited by innovators such as Stability without regard to the efforts and intellectual property rights of creators, then such exploitation will pose an existential threat to those creative industries for generations to come.”
Getty claimed:
Primary infringement of copyright (i.e. that Stability copied Getty’s images for training purposes).
Secondary infringement of copyright (i.e. that the Stability AI model itself is an infringing copy that was imported into the UK).
Trade mark infringement (i.e. that images generated by the Stability AI model reproduced Getty’s watermarks (which are Getty’s trade marks)).
Passing off (i.e. that the reproduction of the watermarks amounted to a misrepresentation in trade that was calculated to injure the business or goodwill of Getty, and caused actual damage to Getty).
Unfortunately the first claim was not pursued because it was accepted that the training did not take place in the UK. It took place in AWS in the US. This is part of why we say above that this particular decision is important but also quite confined/limited.
The decision on secondary infringement turned on what amounts to an “infringing copy” of Getty’s images. What was actually imported into the UK was the AI model itself. Stability argued that copies of the copyright works in question (i.e. the images) were not present within the model. The judge accepted that rather than storing training data, a diffusion model learns the statistics of patterns which are associated with certain concepts found in the text labels applied to their training data i.e. they learn a probability distribution associated with certain concepts. This process of learning the statistical properties of data allows the model to generate new images by sampling from the distribution. So, an AI model which derives or results from a training process involving the exposure of model weights to infringing copies is not an infringing copy itself (refer to para. 600 of the judgment). The secondary infringement claim was dismissed on that basis.
The trade mark infringement claim is probably the most interesting. Getty produced evidence that Stability’s model output images that contained Getty watermarks. For example, when prompted to produce “old Donald Trump behind bars in a jail, news photo” the model output this:
Getty succeeded in the trade mark claim but only in a very limited extent. Due to the stochastic (cf. deterministic) nature of the AI model, it is not possible to determine exactly how often watermarks would be produced, and they would not be produced in the same way every time. The judge found that the watermarks created the impression that there was a material link in the course of trade between Getty and Stability, and that it was possible that an average consumer could be confused as to the source of the output.
The passing off claim was not considered by the court due to the general failure in the trade mark claim.
Interestingly, the judge rejected Stability’s argument that it was the end user and not Stability who was responsible for the model’s output. This means that Stability is not just a “mere conduit” or “passive tool” - it has some responsibility that cannot be completely shifted to end users.
The decision can be found here: https://www.judiciary.uk/wp-content/uploads/2025/11/Getty-Images-v-Stability-AI.pdf