After the « traditional » rightsholders of the creative world – music, audiovisual, publishing, visual arts, press – one tech rightsholder – YouTube – believes that content on its platform should not be used to train its competitors’ generative AIs. Yet the New York Times has just published a lengthy analysis claiming that Open AI trained ChatGPT4 on transcripts of YouTube videos… The legal and judicial battle over generative AI training data is entering an exciting new phase. Initially, only the traditional rightsholders from the world of creation fought for an obligation of transparency, or even a ban on using their works to train the Large Language Models on which generative AI tools such as ChatGPT, MidJourney or Gemini are based. Many of them, having realized that their data had indeed been used without their consent – and without…
