Una società di intelligenza artificiale vince la sentenza dell’Alta Corte dopo la rivendicazione del diritto d’autore da parte dell’agenzia fotografica

The judgment is [here](https://www.bailii.org/ew/cases/EWHC/Ch/2025/2863.html) for those interested (*Getty Images (US) Inc & Ors v Stability AI Ltd*). I wouldn’t be surprised if it gets appealed, given how significant this sort of case might be.

From what I can tell (disclaimer about it being a long, technical judgment that I only skimmed) the key results of the *copyright* side of things (there were also trade mark-related claims) was that while Stability AI did train their models on Getty Images (which would be a breach of copyright) they didn’t do so in the UK, so there was no primary copyright claim.

In terms of images produced based on that trained data (with the same disclaimer as above), Getty couldn’t prove that any particular image generated by the model was a derivative work of any particular Getty image. Which is an interesting problem – the model waters down their images so much, and combines so many together, that it is difficult for Getty to prove copying.

As I understand it some copyright cases elsewhere have been going the other way (or at least being settled), so it will be interesting to see what happens here. If this ruling stands, and I’m reading it correctly, provided people train their models outside the UK they can use as much copyrighted content as they like (and the more the better, to water specific images down).