COURT-RULING

Getty v Stability AI

Updated 2 May 2025
court-rulingcopyrightstability-aiukimage-generation

Getty v Stability AI

Getty Images — the world’s largest commercial photo library — is suing Stability AI for training Stable Diffusion on millions of Getty’s copyrighted photographs. The case tests UK copyright law against AI image generation.

Where NYT v OpenAI tests text, this case tests images. Together, they’ll define whether AI companies need to license the data they train on.

The Facts

Getty Images alleges that Stability AI:

  • Scraped over 12 million Getty images (with watermarks) to train Stable Diffusion
  • Did not obtain a licence or pay for this use
  • Stable Diffusion can generate images in Getty’s style, sometimes even reproducing the Getty watermark
  • This infringes Getty’s copyright and database rights

The watermark reproduction is particularly damning — it demonstrates that the model memorised specific images rather than just learning general visual concepts.

The Question

Does training an AI image model on copyrighted photographs, and generating new images based on that training, infringe copyright under UK law?

UK copyright law doesn’t have the US “fair use” doctrine. Instead, it has narrower “fair dealing” exceptions for research, private study, criticism, and reporting. Whether AI training fits any of these exceptions is untested.

There’s also the question of database rights — a specifically European right that protects substantial investment in collecting data. Getty argues Stability AI extracted a substantial part of their database.

Where It Stands

  • Filed in the UK High Court in January 2023
  • Getty also filed a parallel case in the US (Delaware)
  • UK proceedings are ongoing; this is expected to be a landmark ruling
  • Stability AI has faced financial difficulties, which may affect the case trajectory

Why It Matters

For image generators: All current image generation models (DALL-E, Midjourney, Stable Diffusion, Flux) were trained on scraped images. The legal status of this training is unresolved. This case provides an answer for UK jurisdiction.

For photographers and artists: If Getty wins, it establishes that creators have rights over AI training use of their work. This could lead to licensing frameworks or compensation mechanisms.

For the UK AI ecosystem: The UK has positioned itself as AI-friendly. How its courts handle this case signals whether that friendliness extends to training data practices.

Connection to EU law: The EU AI Act requires transparency about training data. The EU copyright directive also has provisions for text and data mining. This case, while UK, will influence EU thinking. See GDPR & AI for the data protection angle.

Go Deeper

Sources

enes