Rights & provenance
Getty v. Stability, explained
Last updated 2026-07-22
Getty Images sued Stability AI in the UK High Court over the training and outputs of Stable Diffusion. The November 4, 2025 judgment ([2025] EWHC 2863 (Ch)) was the first UK ruling on generative-AI training and copyright.
Why it matters to a buyer
Headlines framed this case both ways; the judgment is narrower than either framing. Getty largely lost, but not because training was ruled lawful — the training claims were never decided, because the training happened outside the UK. The operative lesson is jurisdictional: where a model is trained determines which copyright law applies, and that is a fact to establish in diligence, not assume.
Why it matters to a data owner
The case did not weaken owners’ rights over training; it showed how hard those rights are to enforce across borders. A signed licence removes the enforcement problem by putting terms on paper before anything moves.
Current legal status
UK High Court, November 4, 2025, [2025] EWHC 2863 (Ch). Getty withdrew its primary training and output claims mid-trial because the training occurred outside the UK, leaving the court without jurisdiction over them. Getty lost on secondary infringement — a model that stores no copies is not an "infringing copy" under s.27 CDPA — and won only narrow trademark findings tied to watermarks in generated outputs. It is not a rights-holder win on training, and it decided nothing about whether training itself infringes. Getty was granted leave to appeal on December 16, 2025. The parallel US case, voluntarily dismissed in Delaware and refiled in the Northern District of California in August 2025 (3:25-cv-06891), continues with copyright, DMCA, and trademark claims; on April 23, 2026 the court let the trademark claims proceed while dismissing the DMCA false-CMI claim, and the copyright claim was not part of that motion.
What fiund does about it
A signed licence does not leave training rights to a jurisdictional fight. fiund papers the grant at the source, with provenance records available in diligence.
Sources
- Judiciary of England and Wales — Getty Images v Stability AI, [2025] EWHC 2863 (Ch) (judgment PDF)
- Latham & Watkins — English High Court rejects secondary copyright claim
- Ropes & Gray — Getty loses copyright claim against Stability AI
- Loeb & Loeb — N.D. Cal. order of April 23, 2026 (trademark claims proceed; false-CMI claim dismissed)
Frequently asked questions
Did Getty win or lose in the UK?
Getty largely lost — but not because training was ruled lawful. It withdrew its primary training and output claims mid-trial because the training occurred outside the UK, lost on secondary infringement, and won only narrow trademark findings tied to watermarks in generated outputs.
Did the UK judgment decide whether AI training infringes copyright?
No. The training claims were never decided. The operative lesson is jurisdictional: where a model is trained determines which copyright law applies — a fact to establish in diligence, not assume.
Is the case over?
No. Getty was granted leave to appeal in the UK on December 16, 2025, and the parallel US case, refiled in the Northern District of California in August 2025, continues — on April 23, 2026 the court let the trademark claims proceed while dismissing the DMCA false-CMI claim, and the copyright claim was not part of that motion.
More on rights & provenance
Want data that clears this in diligence?
Whether you're building a model or sitting on an archive, the first conversation is short and specific.
Send a brief