There’s a kaleidoscope of copyright cases aimed at setting boundaries on what AI companies can and can’t do with human-produced creative work. Like with other decisions, a recent ruling in Getty’s AI copyright case could impact what AI tools are allowed to serve up to their users.
In a London-based case brought against Stability AI by Getty Images, Justice Joanna Smith ruled on Tuesday that the AI company, which makes the popular Stable Diffusion image models, did not violate copyright law in training those models. Smith said that Stability AI did not infringe upon Getty’s image copyright protections because it doesn’t “store or reproduce any Copyright Works and nor has it ever done so.”
As with many AI lawsuits, the UK court decision was narrow and nuanced rather than sweeping. Smith determined that Getty succeeded “in part” when arguing that Stability AI had violated its trademark protections by allowing its users to create images that resemble the iStock and Getty Images logos. That success, she said, applies only under certain statutes or laws.
Don’t miss any of our unbiased tech content and lab-based reviews. Add as a preferred Google source.
Smith called her findings both “historic” and “extremely limited” in scope. It’s a sentiment that echoes rulings issued by US courts, highlighting the lack of a consensus among judges when it comes to dealing with copyright claims in the age of AI.
The UK lawsuit was one of the first big cases involving a major content library, which alleged that an AI company had acted illegally by scraping its content from the web. Companies like Stability AI need a vast quantity of human-generated content to build their models. In cases involving similar allegations in the US, Anthropic and Meta emerged largely victorious over authors claiming their books were used for training AI data models without their permission or compensation.
Because of the complexities in Tuesday’s ruling, both companies found room to claim victory.
Getty called the outcome a win for intellectual property owners, given that the ruling said Stable Diffusion infringed Getty trademarks when it included them in AI‑generated outputs.
Watch this: The Hidden Impact of the AI Data Center Boom
“Crucially, the Court rejected Stability AI’s attempt to hold the user responsible for that infringement, confirming that responsibility for the presence of such trademarks lies with the model provider, who has control over the images used to train the model,” Getty said in a statement.
However, Smith’s ruling addressed secondary copyright claims presented by Getty after it dropped its primary claims earlier this year, a point that Stability AI focused on.
“Getty’s decision to voluntarily dismiss most of its copyright claims at the conclusion of trial testimony left only a subset of claims before the court,” Stability AI’s general counsel Christian Dowell said in a statement, “and this final ruling ultimately resolves the copyright concerns that were the core issue.”
Smith stressed that her ruling is specific to the evidence and arguments presented in this particular case. That means that another, similar case could have a different outcome, depending on the exact claim and statute being considered. Similar legal intricacies have been at play in other copyright infringement rulings.
US copyright law has a decades-long history of precedent and a four-part test for judges to consider. However, the novelty of generative AI technology has presented a number of questions for courts to consider, with advocates arguing that existing law is insufficient to protect creators.
Every ruling we receive from these cases helps establish a new set of precedents for courts to consider. For creators, this new ruling means two things. For one, those who use Stability AI in the UK should be able to continue doing so unimpeded. Yet creators who worry about their work being used to train AI models still face the possibility of having their digital content included in training databases.
