Getty Images sued Stability AI, the company behind Stable Diffusion, alleging that training the model on millions of Getty photographs without a licence and without remuneration amounted to mass copyright infringement.
‍

The dispute raises two separate and essential questions. First, does using protected works to train AI amount to reproduction requiring authorisation? Second, who owns copyright in AI-generated images?
‍

As to the first question, U.S. and European courts are still at an early stage of analysis. The principal defence of AI companies is that training is a transformative use and that the model does not 'store' the original images.
‍

As to rights in the AI output, most major legal systems currently do not recognise copyright in works generated exclusively by AI, without human creative contribution.

Companies using AI to generate commercial content must therefore be aware that they may own limited rights or no rights at all in that content.