Getty Drops Key Copyright Claims Against Stability AI, But UK Lawsuit Remains Active
AI-generated images have been under legal scrutiny for years, especially when models like Stable Diffusion are trained on copyrighted material. One of the biggest questions asked by creators, legal experts, and developers is: Can AI companies legally use copyrighted content to train their models? That’s the focus of Getty’s lawsuit against Stability AI — and now, a significant update has changed the conversation. Getty has dropped its primary copyright claims in a London court, reshaping one of the most high-profile AI-related lawsuits. However, other legal challenges remain, both in the UK and the US.
UK Copyright Lawsuit: Getty Narrows Focus on Stability AI
Getty Images originally filed a lawsuit against Stability AI in January 2023, claiming the company behind Stable Diffusion had used millions of copyrighted images — many of them from Getty’s own library — to train its generative AI model without permission. Getty also alleged that AI-generated outputs sometimes resembled copyrighted works or included Getty’s watermark, implying misuse.
As of June 2025, Getty dropped both of those primary copyright infringement claims in the UK. This decision signals a major shift in the case's trajectory and highlights the challenges of enforcing traditional copyright laws on emerging generative AI technologies. According to legal experts, the dropped claims likely stemmed from insufficient jurisdictional evidence and weak arguments tying the AI outputs directly to substantial copyrighted content.
While this move marks a partial win for Stability AI, it doesn’t spell the end of legal risks for AI model developers, especially those using content scraped or borrowed from online sources without licensing.
Secondary Infringement and Trademark Claims Still Stand
Getty’s legal team may have dropped the core copyright complaints, but they aren’t walking away from the case entirely. The lawsuit still includes claims of secondary infringement and trademark violations. These remaining claims suggest a new legal strategy: even if AI training happened outside the UK, importing and using those models inside UK borders may still breach copyright laws.
Legal experts believe this approach could become a benchmark for similar international disputes. It raises questions about whether an AI model itself — not just its training data — could be treated as an "infringing article" under UK law. That could have wide-reaching consequences for global AI startups that build their models in one country and deploy them in another.
Meanwhile, Getty is also pursuing a related lawsuit in the United States, where it alleges that as many as 12 million copyrighted images were used without consent to train Stability AI’s models. That case, which includes both copyright and trademark claims, is progressing independently and could potentially influence how U.S. courts treat generative AI systems.
What Getty’s Move Means for AI Developers and Copyright Holders
Dropping the UK copyright training and output claims shows how complex — and legally untested — the issue of AI training on copyrighted material remains. Getty’s decision highlights how difficult it is to prove that AI-generated images directly infringe on original copyrighted works, especially when the model’s outputs are unpredictable and vary from the training data.
This has wider implications for AI startups, model trainers, and platform providers. First, companies will need to pay closer attention to jurisdictional nuances when training and deploying AI models. Second, this case reinforces that transparency and documentation around training data could become key legal defenses or liabilities. Lastly, the evolution of the UK and U.S. lawsuits could shape new guidelines for what constitutes fair use, copyright infringement, and secondary liability in the AI era.
For copyright holders, Getty’s partial legal retreat serves as both a warning and a strategy playbook. Pursuing large-scale copyright claims may be harder to win under current legal frameworks. However, trademark claims and secondary infringement arguments tied to the distribution of trained AI models could offer stronger grounds for legal action.
Stability AI’s Response and Broader Industry Reactions
Stability AI, unsurprisingly, welcomed Getty’s decision to narrow the UK lawsuit. In a public statement, the company said it was pleased to see multiple claims dropped after witness testimonies concluded. Stability also pushed back against remaining trademark claims, arguing that users don’t interpret any residual watermarks in AI outputs as endorsements or branding from Stability itself.
The outcome of this case — still very much in motion — is being closely watched by other companies in the generative AI space, such as OpenAI, Midjourney, Runway, and Anthropic. Just one day before Getty’s update, a U.S. judge sided with Anthropic in a similar lawsuit over training AI on copyrighted books, offering another legal datapoint in this fast-evolving arena.
While courts haven’t yet delivered a definitive answer on whether training AI on copyrighted content constitutes infringement, one thing is clear: companies will need to tread carefully. Legal strategies are becoming more nuanced, moving beyond simple questions of data usage into more layered discussions about importation, model liability, and downstream effects.
Post a Comment