Key Takeaways
- A dataset search tool created by The Atlantic revealed that millions of creative works—including songs by Australian artists such as Paul Dempsey, Bernard Fanning, Kylie Minogue, and Darren Hayes—have been scraped from the internet to train AI models.
- Artists argue that the unlicensed use of their material in AI training datasets violates contractual agreements, undermines fair compensation, and dehumanises the creative process.
- The controversy has sparked calls for stronger copyright protections, with organisations like APRA AMCOS urging tech companies to negotiate licensing terms rather than rely on “text and data mining” exemptions.
- Although the Australian government rejected a Productivity Commission proposal that would have legalised unpaid AI training on copyrighted content, the debate over “slurping” of creative work remains unresolved.
Introduction to the Controversy
The recent release of a dataset search tool by US publication The Atlantic has shone a spotlight on the widespread practice of harvesting creative content from the internet to train artificial‑intelligence models. The tool indexes millions of tracks, lyrics, and texts that have been scraped from platforms such as YouTube and Genius.com, revealing that a substantial portion of Australia’s musical and literary catalogue is embedded in these training sets. For many musicians, the discovery confirms long‑standing suspicions that their work is being used without consent or compensation, prompting a public outcry that blends legal, ethical, and artistic concerns.
Paul Dempsey’s Reaction and Quote
Paul Dempsey, frontman of the acclaimed band Something For Kate, described the situation as “frustrating” and said he felt his lifelong negotiations with record labels and other entities had been rendered useless. In an interview with AAP, he stated, “It’s frustrating this is happening. Every negotiated agreement and contract I’ve ever gone into in my career with whatever entity or record label is all just rendered useless.” Dempsey added, “An artist’s ability to negotiate fair terms for the use of their content is just being ripped away from them.” His words capture the sense of betrayal felt by creators who see their intellectual property being repurposed by opaque AI systems.
Bernard Fanning’s Concerns About Dehumanisation
Bernard Fanning, former lead singer of Powderfinger, echoed Dempsey’s frustration while steering the conversation toward the philosophical implications of AI‑generated art. He warned that using original songs to produce robotic output risks stripping away the human essence that makes music meaningful. “Do we want robots telling our stories and synthesising our feelings? Because it’s not human. The whole point of art is to humanise our feelings, to express how we’re feeling across the whole range of emotions,” Fanning told AAP. He concluded bluntly, “Robots aren’t alive; they don’t experience, they just aggregate – and the idea of that sucks.” For Fanning, the threat lies not only in economic loss but in the potential erosion of art’s capacity to connect people on an emotional level.
Darren Hayes’ Instagram Outburst
Pop songwriter Darren Hayes, best known for his work with Savage Garden, took to Instagram to voice his fury after discovering that his three‑decade recording catalogue—including hits like “Truly Madly Deeply”—had been ingested into AI training data. His post read, “I absolutely feel violated that all of the hundreds and hundreds and hundreds of hours, blood, sweat and tears that I’ve put into my music, along with every other musician, has been stolen and served up like french fries to a piece of software that spits out shit.” Hayes’ vivid language underscores the visceral sense of violation many artists experience when their lifelong labour is reduced to training fodder for algorithms they did not authorize.
Details of the Datasets (Sleeping AI, LAION)
The contested material appears in two large‑scale datasets. The first, Sleeping‑DISCO‑9M, assembled by a research group dubbed Sleeping AI, contains 9.7 million music tracks sourced from YouTube alongside lyrics pulled from Genius.com. The second, LAION‑DISCO‑12M, compiled by the Germany‑based organisation LAION, comprises 12.3 million YouTube tracks. While The Atlantic cautioned that the mere presence of a work in these datasets does not definitively prove it was used to train a specific model, the sheer volume of Australian content—spanning pop, rock, and literary works—has intensified scrutiny over how such data is collected and whether proper permissions were ever obtained.
Legal and Industry Perspectives (APRA AMCOS, Productivity Commission)
Music licensing body APRA AMCOS, which represents 128,000 members across Australasia, labelled the datasets as “proof of the theft of creative work.” Its chief executive, Dean Ormston, criticised major tech platforms for avoiding direct negotiations: “Major tech platforms have not come to the table. Not once. Instead, they have lobbied governments, circulated policy papers, and proposed solutions designed to extinguish any obligation to pay.” Australia’s existing intellectual property framework requires permission and agreed‑upon payment before copyrighted works can be used, yet the IT sector has pushed for broad text‑and‑data‑mining exemptions. In August 2025 the Productivity Commission floated reforms that would have legalised AI training on copyrighted material without compensation, but the federal government rejected the proposal in October, signalling a preference to protect creators’ rights—at least for now.
Implications for Artists and Future Outlook
The backlash from Dempsey, Fanning, Hayes, and others highlights a growing tension between technological innovation and creators’ livelihoods. Artists warn that unchecked AI training could flood the market with low‑cost, algorithmically generated content, undermining the value of original work and making it harder for emerging musicians to earn a living. Moreover, the dehumanising critique raises questions about the cultural role of art: if machines can mimic styles without understanding emotion, does the listener’s experience suffer? Moving forward, stakeholders—including legislators, tech companies, and rights organisations—will need to craft licensing models that acknowledge the transformative nature of AI while ensuring fair remuneration and respect for creators’ moral rights.
Conclusion / Closing Thoughts
The revelation that Australian songs and novels are embedded in massive AI training datasets has ignited a vigorous debate about consent, compensation, and the very purpose of artistic expression. As Paul Dempsey lamented, “We can trigger huge emotional responses in each other through art, and I don’t know that that’s going anywhere; it’s just going to be flooded with all this other shit.” Bernard Fanning’s warning about robots telling human stories and Darren Hayes’ visceral sense of violation serve as potent reminders that behind every data point lies a creator’s sweat, passion, and livelihood. Whether the industry will adapt to honour those contributions—or continue to treat them as raw material for machines—remains one of the defining challenges of the AI era.
https://www.theguardian.com/culture/2026/jun/26/australian-musicians-aussie-music-ai-training-tool-nick-cave-kylie-minogue-paul-dempsey