Getty asks London's High Court to stop UK sales of Stable Diffusion, accusing Stability AI of copyright infringement for training its AI model on Getty's images
Sam Tobin / Reuters : Tweets: @mytechmusings Tweets: Prakash Sangam / @mytechmusings : IMO this is only the beginning. #datasovereignty & who has control over data uses for #Ai training will be a major issue going forward... #privacy #Security #ArtificialIntelligence #GanAI https://twitter.com/...
Context & Ripple Effects
Getty's bid for a UK injunction is the second front opened against Stability AI within weeks: the agency had already filed a US lawsuit alleging more than 12 million images were copied without permission in February 2023, after flagging the UK legal action back in January. Unlike OpenAI's closed approach, Stable Diffusion's open-source dataset made it the most exposed target for a rights-holder strike.
The London case ran long and narrow. By the time the High Court trial began in June 2025, Getty had already dropped its primary copyright-infringement claims mid-proceedings, keeping only narrower claims alive alongside the separate US fight — and the court ultimately ruled against Getty in November 2025 with the judge describing the outcome as extremely limited in scope.
First-order effects
- If granted, the injunction halts Stable Diffusion's UK distribution outright — hitting Stability AI's commercial sales and developer access in one of its home-region markets while its US exposure continues through Getty's parallel American suit.
- Stability AI faces legal costs and product-contingency planning across two jurisdictions simultaneously, with the UK injunction request targeting the model itself rather than any single output.
Second-order effects
- Other model builders read the same risk: Stability AI's subsequent release of audio models trained explicitly on licensed data signals the company hedging toward pre-cleared datasets as a defense, effectively pricing licensing into its product roadmap.
- Rights holders gain a template — a injunction-first tactic aimed at distribution rather than damages — which competitors and stock-photo rivals can replicate against open-weight model makers whose training data is public and auditable.
Third-order effects
- The arc from injunction demand to a narrow, limited-scope ruling points to litigation proving a blunt instrument for settling training-data disputes, pushing the industry toward licensing markets and provenance documentation as the practical compliance layer.
- Jurisdiction-by-jurisdiction fights like this one make training-data legality a market-access question: models may end up cleared for some territories and blocked in others depending on how each court reads ingestion versus output infringement.
The trend: Image-rights holders and generative-model makers are contesting training-data legality market by market, and with courtroom outcomes landing narrow, licensed-data pipelines are becoming the industry's de facto settlement.