Australian musicians sound warning note after Nick Cave, Kylie and many more slurped into AI training tool
TL;DR
Paul Dempsey, Bernard Fanning and Darren Hayes are criticizing the discovery of their songs in datasets that can be used to train music AI systems. The Atlantic’s search tool surfaced works by Kylie Minogue, Powderfinger, Nick Cave, Jimmy Barnes, and authors including Thomas Keneally and Peter Carey. The datasets include Sleeping-DISCO-9M, with 9.7 million YouTube tracks plus Genius lyrics, and LAION-DISCO-12M, with 12.3 million YouTube tracks.
Nauti's Take
The excuse is getting thin: music AI builders cannot hide behind dataset fog forever. If YouTube rips and lyrics sit inside training pools, teams need provenance, audit trails, and licensing logic before the next generator turns into a legal construction site.
Briefingshow
This turns the copyright fight around generative music into something concrete: not vague data, but full catalogues by working artists. If scraping bypasses contracts, licences and bargaining power, value shifts from creators to platforms. Australia recently rejected broad text and data mining exemptions, but the tech industry pressure has not gone away.
Sources