AI Giants Harvest Millions of Australian Songs, Dividing Creative Community

Australia's music industry is sounding an alarm over artificial intelligence, with fresh analysis revealing that major tech companies have scraped approximately 47 million songs—including works by some of Prime Minister Anthony Albanese's publicly known musical preferences—to train machine learning models without artist consent.

The figures paint a sobering picture for Sydney's thriving live music scene. According to industry bodies tracking the issue, AI companies have extracted data valued at an estimated $1.8 billion in intellectual property, with Australian artists accounting for roughly 8-12 percent of unauthorised dataset compilations. For context, Sydney venues ranging from the Enmore Theatre in Newtown to smaller rooms across Inner West suburbs depend heavily on recording royalties to sustain operations.

"The maths doesn't work in musicians' favour," explains research collated by independent music advocacy groups. Average artist earnings have declined 23 percent since 2022, precisely when major AI training datasets expanded exponentially. The same data shows that a typical Sydney session musician—common across studios from Marrickville to Alexandria—earns approximately $38,000 annually, down from $49,500 just four years ago.

The scale becomes even starker when examining specific sectors. Independent record labels operating from Australian locations now report that 64 percent of their catalogue has appeared in unauthorised AI training datasets. Manufacturing and distribution costs for physical media have climbed 18 percent, while streaming revenues to artists have stagnated at around $0.003 per stream.

Federal Parliament's 47 Australian seats with significant creative constituencies face constituent pressure on this issue. Organisations representing Sydney's music community—from venues along King Street, Newtown to recording facilities across the CBD—have collectively raised over $2.4 million through crowdfunding to pursue legal challenges since 2024.

Government response has so far proven measured. Current legislation provides minimal protection, with copyright enforcement mechanisms designed decades before AI's emergence. Australian music exports, valued at $368 million in 2024-25, face potential erosion if training practices continue unchecked.

The Prime Minister's office has acknowledged the crescendo of concern from the creative sector. However, translating concern into policy remains complex—particularly when balancing innovation incentives against artist protections. Industry modelling suggests that comprehensive AI governance could preserve up to 340 jobs across Sydney's music production sector within three years.

As the debate intensifies, one statistic stands out: 89 percent of Australian musicians surveyed support mandatory licensing agreements for AI training data. Whether Parliament acts on this mandate before 2027 will define a generation of Australian music's digital future.

This article was compiled by AI and screened before publishing. See our editorial standards.