Australian and New Zealand Music Found in AI Training Datasets Without Consent
An investigation by The Atlantic identified millions of Australian and New Zealand songs in AI training datasets shared within the AI development community without artist consent, licensing, or payment. APRA AMCOS, representing 128,000 songwriters and composers, estimates creators could lose over $500 million in royalties over four years without mandatory licensing frameworks.
What happened
An investigation by The Atlantic uncovered millions of songs including Australian and New Zealand musical works in four 'giant datasets of songs that are being shared within the AI-development community' used for AI training without consent, licensing, or payment.
The Atlantic's AI Watchdog tool allows anyone to search and identify whether artists' tracks have been included in AI training systems. APRA AMCOS, representing 128,000 songwriter, composer and publisher members, identified songs from artists including Kylie Minogue, Midnight Oil, Cold Chisel, INXS, Sia, Crowded House, Flume, and Nick Cave, among many others.
Reported impact
APRA AMCOS estimates that without a mandatory licensing framework, creators across Australia and New Zealand could miss out on more than $500 million in royalties over the next four years.
According to APRA AMCOS, 940 Kylie Minogue songs have been identified in a single AI dataset, including tracks such as 'Spinning Around', 'Can't Get You Out of My Head' and 'All the Lovers'. Nick Cave's novel 'The Death of Bunny Munro' has also reportedly been incorporated into datasets.
Governance controls
AI companies have escalated lobbying of the Australian and New Zealand Governments to reopen copyright law as part of negotiations over major data centre investment commitments.
APRA AMCOS chief executive David Ormston stated that technology companies have failed to engage meaningfully with creators and instead have lobbied governments, circulated policy papers, and proposed solutions designed to extinguish any obligation to pay.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.