NLJ this week: Failure to deliver in Getty v Stability AI

The High Court found limited trade mark infringement where AI-generated images reproduced Getty watermarks, but rejected broader claims of dilution, passing off, and secondary copyright infringement. Crucially, the court held that AI models are not, in themselves, infringing copies of their training data.

The article explains why Getty’s evidential hurdles proved decisive: contrived prompts and small samples were not enough to show real-world infringement in the UK.

The authors argue the case sets a demanding standard for future AI claims, requiring robust, jurisdiction-specific evidence tied to actual user behaviour. For rights-holders and developers alike, the lesson is clear—speculation will not substitute for proof in the AI age.