Getty Images dropped its primary claims of copyright infringement against Stability AI on Wednesday at London’s High Court, narrowing one of the most closely watched legal fights over how AI companies use copyrighted content to train their models.
The move doesn’t end the case entirely – Getty is still pursuing other claims as well as a separate lawsuit in the U.S. – but it underscores the gray areas surrounding the future of content ownership and usage in the age of generative AI. The development also comes just a day after a U.S. judge sided with Anthropic in a similar dispute over whether training AI on books without author permission violates copyright law.
Getty sued Stability AI — the startup behind AI image generator Stable Diffusion — in January 2023 after alleging that Stability used millions of copyrighted images to train its AI model without permission.
The image database company also claimed that many of the works generated by Stable Diffusion were similar to the copyrighted content used to train it. Some, Getty said, even had its watermarks on them.
Both of those claims were dropped as of Wednesday morning.
“The training claim has likely been dropped due to Getty failing to establish a sufficient connection between the infringing acts and the UK jurisdiction for copyright law to bite,” Ben Maling, a partner at law firm EIP, told TechCrunch in an email. “Meanwhile, the output claim has likely been dropped due to Getty failing to establish that what the models reproduced reflects a substantial part of what was created in the images (e.g. by a photographer).”
In Getty’s closing arguments, the company’s lawyers said they dropped those claims due to weak evidence and a lack of knowledgeable witnesses from Stability AI. The company framed the move as strategic, allowing both it and the court to focus on what Getty believes are stronger and more winnable allegations.
What remains in Getty’s lawsuit are a secondary infringement claim as well as claims for trademark infringement. Regarding the secondary infringement claim, Getty is essentially arguing that the AI models themselves might infringe copyright law, and that using these models in the UK could constitute importing infringing articles, even if the training happened outside the UK.
“Secondary infringement is the one with widest relevance to genAI companies training outside of the UK, namely via the models themselves potentially being ‘infringing articles’ that are subsequently imported into the UK,” Maling said.
A spokesperson for Stability AI told TechCrunch the startup was “pleased to see Getty’s decision to drop multiple claims after the conclusion of the testimony.”
The spokesperson also noted that Stability was confident that Getty’s trademark and passing off claims will fail because consumers don’t interpret the watermarks as a commercial message from Stability AI.
Getty’s U.S. division also sued Stability AI in February 2023 for trademark and copyright infringement. In that case, Getty alleged that Stability used as many as 12 million copyrighted images to train its AI model without permission. The company is seeking damages for 11,383 works at $150,000 per infringement, which would amount to a total of $1.7 billion.
Separately, Stability AI is also named in another complaint alongside Midjourney and DeviantArt after a group of visual artists sued the three companies for copyright infringement.
Getty Images has its own generative AI offering that leverages AI models trained on Getty iStock stock photography and video libraries. The tool allows users to generate new licensable images and artwork.