Stability AI wins UK court battle against Getty Images (apnews.com)

Stability AI largely prevailed in a High Court ruling in London after Getty Images accused the company of scraping roughly 12 million Getty photos to train its Stable Diffusion image generator. Justice Joanna Smith rejected Getty’s remaining secondary copyright claims, finding that Stable Diffusion “does not store or reproduce any Copyright Works (and has never done so),” while narrowly upholding a limited trademark claim where Getty watermarks appeared on some generated images. Getty had dropped its primary copyright allegation during the trial and also argued that training occurred outside the UK (on Amazon infrastructure), a point that undercut its jurisdictional theory. The decision is significant because it preserves important room for AI developers to argue that model training on large image corpora does not necessarily amount to copyright reproduction, but it leaves the legal landscape unsettled. The court explicitly declined to decide broader questions abandoned by Getty, so core issues about the lawfulness of training on copyrighted material remain unresolved in the UK. The ruling also highlights practical risks for providers: isolated instances of watermark reproduction can trigger trademark liability. Getty has filed a parallel U.S. suit, and the case sits amid dozens of global lawsuits that will collectively shape future limits on dataset use, model provenance obligations, and how “fair dealing/fair use” applies to generative AI.