Australian Musicians Slam AI Training Tool for Using Their Songs Without Permission

A growing number of Australian musicians are speaking out against the use of their original songs in artificial intelligence training datasets. The controversy surrounds the inclusion of millions of creative works, including songs by Kylie Minogue, Nick Cave, and Jimmy Barnes, in datasets used to train AI models. Paul Dempsey and Bernard Fanning are among the musicians who have expressed frustration and concern over the practice, which they say undermines their ability to negotiate fair terms for the use of their content. Dempsey, whose music with Something For Kate and as a solo artist is included in the datasets, believes that the use of his songs without permission renders his contracts and agreements useless. Bernard Fanning argues that using original songs to produce AI-generated content is dehumanizing and goes against the purpose of art, which is to express human emotions. Other musicians, including Darren Hayes, have also spoken out against the practice, with Hayes expressing his fury on Instagram over the inclusion of his entire 30-year recording career in the datasets. The datasets in question, Sleeping-DISCO-9M and LAION-DISCO-12M, were assembled by researchers and contain millions of music tracks from YouTube, as well as lyrics from Genius.com. While the inclusion of songs in these datasets does not necessarily mean they have been used to train AI models, music licensing organization APRA AMCOS sees it as proof of the theft of creative work. The organization's chief executive, Dean Ormston, criticizes major tech platforms for not coming to the table to discuss payment and terms for the use of copyright works. The issue highlights the ongoing debate over intellectual property laws and the use of content in AI training, with the Australian government having recently ruled out changes that would have legalized the use of content without paying creators.