Three American giants are actively protecting their intellectual property rights against generative AI, as two legal battles commence on both sides of the Atlantic. In the UK, Seattle-based media company Getty Images accuses UK-based Stability AI of multiple IP infringements. In the US, The Walt Disney Company and Universal Studios are teaming up against Midjourney, an AI startup, with their main ground being copyright infringement.
Both cases are centered around questions legal minds have been posing since the introduction of generative AI: Is the output of generative AI an infringement? And who is ultimately responsible for the output, the platform or the user?
Getty Images v. Stability AI
Getty initially filed a claim in the High Court in 2023, which resulted in Stability applying for reverse summary judgment on the grounds that Getty had no real prospect of success, arguing that their operations took place outside the UK. However, the High Court judge hearing the case decided that the claims brought by Getty did have a real prospect of succeeding in court. Despite this, Stability saw a small victory when the court ruled that the representative action brought by Getty would not succeed due to the difficulties in identifying who qualified for the class. The proposed class was comprised of 50,000 rightsholders who alleged their rights were also infringed. Stability was successful in arguing that identifying these individuals would be challenging due to the unclear definition of the class.
This current trial is centered around four main grounds:
- Copyright infringement. Getty accuses Stability of using content that Getty owns or has an exclusive license for when training their model, Stable Diffusion, resulting in the generated output containing substantial parts of that content. Getty is also alleging secondary copyright infringement, arguing that Stability is importing an article into the UK that they know is infringing Getty’s content (i.e., Stable Diffusion). During the trial’s closing speeches, Getty dropped its copyright infringement claim to focus only on the remaining claims—likely due to challenges producing sufficient evidence to satisfy the threshold under UK law. For its part, Stability argues that the training took place outside of the UK and there were no reproductions made of Getty’s copyright within the UK. As such, they argue, there is no secondary copyright infringement, and any output that features Getty’s images was purposefully created by Getty and generated by using prompts in a way that does not reflect how the average user would use the technology.
- Trademark Getty accuses Stability of using Getty’s watermarks to train its AI models and that these watermarks are included in the final output. Stability argues that any output showing Getty’s watermarks was created by Getty’s own legal team and was not as a result of using Stability in the ordinary course of trade. They also argue that any watermarks seen on outputs were not used by Stability and do not result in confusion or unfair advantage.
- Database rights infringement. Getty argues that Stability has infringed their rights by using a large part of their database and reusing it for their own purposes. This argument is based on their claim that their large content collection should be deemed a database under the Copyright and Rights in Database Regulations 1997. Stability argues that Getty’s collection does not meet the requirements to be classified as a database and should not be afforded the associated protection. Additionally, the alleged extraction did not occur within the UK, and they have not reused the content, as there is no correlation between the output and Getty’s content.
- Passing off. Getty argues that outputs use content from Getty’s Getty Images and iStock, resulting in the misrepresentation that the output is owned or licensed by Getty. Stability argues that (1) any such misrepresentation claimed would be made by the user and not by Stability, and (2) any similarity in the output is not enough to confuse or deceive the general public.
The trial in the UK began on June 9, 2025, and is due to end on June 30, 2025. The judgment is expected sometime in July.
Disney and Universal v. Midjourney
Before suing the generative AI company Midjourney, which specializes in text-to-image generation, both Disney and Universal sent separate cease-and-desist letters that reportedly went unanswered. Following this, the two media companies jointly filed a case in the US District Court of Los Angeles.
The primary arguments being put forward by Disney and Universal are also copyright infringement and secondary copyright infringement, echoing those being made by Getty in its case against Stability. It is too early to know what Midjourney will argue in their defense, but similar arguments to Stability AI are expected—that the users are the ones responsible for the outputs.
What Does This All Mean?
Although these cases are taking place in two different jurisdictions, the impact of their judgments is likely to be felt globally. With generative AI still in its infancy, many jurisdictions are still not well equipped to deal with the legal ramifications of content being created by this ever-evolving technology. As a result, many legal systems are looking at others before determining their next steps.
The long-employed defense of fair use will be put to the test yet again, and the lines may be drawn more clearly with a potential decline in generative AI companies being able to rely on it. This defense has already been rejected in one US court in the recent decision from the Thomson Reuters v. Ross Intelligence case. However, the more recent decision in Bartz v. Anthropic resulted in a federal judge allowing the defense of fair use and ruling that training the AI model on books without the author’s consent is not infringement. It is worth noting that in that case, the books were obtained from pirate websites, and so there will be another trial to determine the legality of the library Anthropic used.
Although the courts seem to be split over whether fair use should be allowed in relation to generative AI, the decisions do suggest that the source material should be legally obtained prior to training the models. Therefore, AI companies may need to start looking for other solutions to prevent future claims of infringement, such as licensing content for AI model training. This move would be welcomed by rightsholders but would result in additional costs for the AI startups.
Governments are likely to feel more pressure to put in place AI laws that bring more clarity than current laws and regulations on AI—especially in the UK, now that Getty has dropped their copyright infringement case and there will be no imminent judgment on this issue from the High Court. With the fast-paced nature of generative AI, it is likely governments will find themselves in an increasingly difficult position, with some wanting to keep a more laissez-faire approach and others wanting more stringent protections for their intellectual property. In the absence of court decisions to bridge the gap, many common-law jurisdictions will now be looking for legislative changes instead.
The decisions coming from the UK and US courts may be highly persuasive around the world and could lead to legislative changes to bring some stability for rightsholders and users. After all, IP laws tend toward global alignment, and it is unlikely rightsholders or governments will want to go against the stream in such uncertain times.
