As a former head of audio at Stability AI, Newton-Rex is well acquainted with the internal data training practices of generative AI companies. He resigned over the company's belief that training their AI models on copyrighted content without licensing it constitutes "fair use."

Generative AI models have generally scrapped as much content as they could from the internet, downloaded it and trained their model to be able to create new work that is in the style of the work it trained on, Newton-Rex told ABC News.