Getty v. Stability AI: Trademark and Copyright Lessons

What the two Getty cases are about

Getty Images (US), Inc. and related companies sued Stability AI, the company behind the Stable Diffusion image generator, in the UK and in the United States. The two cases are separate. The UK case applies UK law, including the Copyright, Designs and Patents Act 1988 (CDPA) and the Trade Marks Act 1994. The U.S. case applies U.S. law, including the Copyright Act and the Lanham Act.

The UK trial concerned two questions. The first was whether Stable Diffusion’s model weights are an “infringing copy” of Getty’s photographs. The second was whether normal use of the models in the UK produces images that carry Getty Images or iStock watermarks.

UK ruling of 4 November 2025: copyright

The judgment is Getty Images (US) Inc & Ors v Stability AI Ltd [2025] EWHC 2863 (Ch), by Mrs Justice Joanna Smith.

Getty dropped two claims before the end of the trial. Its training-and-development claim failed on the facts because there was no evidence that training took place in the UK. Its claim about generated outputs ended because Stability AI had blocked the prompts used, so the relief sought had substantially been achieved. The database-right claim could not go on without them. On 16 December 2025 the court’s order recorded the abandoned primary-copyright and database-right claims as dismissed.

That left a claim of secondary infringement under sections 22 and 23 of the CDPA. Getty argued that the model was an “infringing copy” because making its weights would have infringed copyright if done in the UK. The court accepted that an “article” can be intangible, but held that Stable Diffusion “does not store or reproduce any Copyright Works (and has never done so).” It dismissed the claim.

UK ruling: trade marks and watermarks

Getty’s trade mark claim asked whether normal UK use of Stable Diffusion produces images with Getty Images or iStock watermarks. The court found infringement only in part. It said it was “impossible to know how many (or even on what scale)” such watermarks were generated in real life. In a costs ruling of 17 December 2025, the judge called Getty’s trade mark findings “historic and extremely limited in scope.”

The findings, model by model:

• iStock watermarks from v1.x models: infringement under section 10(1) and section 10(2) of the Trade Marks Act 1994, based on two example images.
• Getty Images watermarks from v2.x models: infringement under section 10(2), based on one example image. The court found no infringement of the Getty Images marks for v1.x models.
• Section 10(3) (marks with a reputation): dismissed.
• Stable Diffusion XL and v1.6: the claim failed because there was no evidence that a UK user generated either watermark with them.
• Passing off: the court declined to decide it.

What happens next in the UK

On 16 December 2025 the judge gave Getty permission to appeal on the secondary-infringement point. She called it a point of law that is “both novel and important” and said the appeal has a real prospect of success. She refused Stability AI permission to appeal the trade mark findings, because its grounds challenged findings of fact.

As of 26 October 2026, the National Archives’ Find Case Law service listed no Court of Appeal judgment in this case after the High Court rulings of December 2025.

The U.S. case in California

Getty’s U.S. case is Getty Images (US), Inc. v. Stability AI, Ltd., No. 3:25-cv-06891 (N.D. Cal.), before Judge Trina L. Thompson. It was filed on 14 August 2025. Getty had earlier sued in the District of Delaware (No. 1:23-cv-00135).

The docket shows an order on Stability AI’s motion to dismiss on 23 April 2026 (ECF 53), Stability AI’s answer on 21 May 2026 (ECF 62) and an amended answer on 2 July 2026 (ECF 67). Filings continued through 4 September 2026. The docket entries read for this guide show no ruling on fair use.

Fair use is decided under 17 U.S.C. 107. It weighs four factors: the purpose and character of the use, the nature of the work, the amount and substantiality of the portion used, and the effect on the market for the work. In May 2025 the U.S. Copyright Office published a pre-publication report on generative AI training and said a final version will follow.

What this means for a business that uses AI image tools

The UK copyright holding is about whether a model counts as an infringing copy under UK law. It does not say that every AI output is safe to publish.

The trade mark findings point to a separate risk. An image that carries a distorted third-party watermark or logo can suggest a connection with that company. In the U.S., the Lanham Act covers use in commerce of a copy or imitation of a registered mark that is likely to cause confusion (15 U.S.C. 1114(1)). It also covers a false designation of origin that is likely to cause confusion about affiliation, sponsorship or approval (15 U.S.C. 1125(a)(1)(A)). Steps a business can take:

• Check generated images for logos, watermarks and brand names before you publish them.
• Keep a record of the tool, the prompt and the date for images used in advertising or on products.
• Read the AI tool’s terms on who owns the output and who is responsible for it.
• Ask a lawyer before you build a product or campaign on a model whose training data you cannot check.