Kylie Minogue has become the latest high-profile artist drawn into a widening fight over artificial intelligence and copyright, after a music-publishing industry group said AI models have already learned from her songs without payment or permission. The claim places the Australian pop star alongside Nick Cave in a debate that is rapidly moving from abstract tech policy to the careers and royalties of individual musicians.
According to the International Confederation of Music Publishers, or ICMP, AI systems have trained on music by Minogue and Cave, and neither artist saw a cent for it. The allegation is not a narrow complaint about one platform or one model. It is part of a broader warning that creative work is being ingested at industrial scale while lawmakers debate whether to weaken the rules that are supposed to protect it.
Headline and key facts
Headline: Kylie Minogue is the latest artist caught in the wholesale AI theft storm.
- Kylie Minogue and Nick Cave have been named by the ICMP as artists whose songs were used to train AI models.
- The ICMP says neither artist was paid for that use.
- ICMP head John Phelan accused AI companies of asking lawmakers to “legitimise larceny.”
- Tech firms have reportedly asked dozens of governments to relax copyright rules for AI training.
- Australia appears to be the only country giving the request a second look.
- Prime Minister Anthony Albanese said in July that it amounts to “theft” if creators cannot decide whether AI learns from their work or how much they earn from it.
- A parliamentary AI committee has been described as deadlocked on the issue, with lawmakers from both sides raising concerns.
- AI companies have argued that the most advanced models will not be trained in Australia unless local copyright laws change.
Why Kylie Minogue’s catalogue matters in the AI debate
Minogue is not a random name attached to a technical policy dispute. She is one of Australia’s most recognisable cultural exports, with a career that stretches from soap opera and 1980s pop through to dance-floor reinvention and a recent resurgence. Her catalogue includes songs that have been streamed, licensed, covered, sampled and performed around the world for decades. That long commercial life is exactly what makes her music valuable to AI developers building models that can generate lyrics, melodies, production styles and synthetic voices.
For publishers and songwriters, the concern is not only that a model might imitate a particular singer. It is that the underlying compositions, lyrics and recordings can be used as raw material to build a competing commercial product. If those works are ingested without licences, tracking or payment, the harm is both economic and creative. Artists lose leverage over how their work is used, while audiences may hear outputs that trade on a lifetime of recognisable style without any direct connection to the people who made it.
Minogue’s global profile gives the argument a face. A pop star with a multi-decade catalogue can at least attract public attention and legal resources. Lesser-known songwriters, session musicians and independent publishers often have far less ability to detect whether their work has been swept into a training dataset, let alone negotiate a fee. That imbalance is central to the industry’s complaint: AI systems are built by some of the largest companies in the world, while the rights holders whose work feeds them are fragmented across thousands of individual contracts and territories.
Nick Cave and the broader warning from music publishers
Nick Cave brings a different kind of weight to the issue. As one of Australia’s most respected singer-songwriters, he is associated with a body of work known for narrative lyrics, literary ambition and a distinctive artistic identity. His inclusion in the ICMP’s warning underlines that the debate is not confined to glossy pop hits. It reaches across genres, generations and reputations. If AI models can learn from Cave’s songs without consent or compensation, the same principle applies to poets, composers, producers and emerging artists who have even less bargaining power.
The ICMP represents music publishers, the companies and organisations that manage the rights to songs and compositions. Its intervention matters because copyright in music is layered. A single track can involve separate rights in the sound recording, the underlying composition, the lyrics and the performance. AI training can touch all of those layers. A model that learns patterns from lyrics may not reproduce a specific line, but it can still absorb and imitate stylistic traits that were developed over a career. Publishers argue that this is a commercial use of creative work and should be licensed, just as radio, streaming services and sync placements are licensed.
The accusation of “legitimise larceny”
John Phelan, who leads the ICMP, did not soften his language. He accused AI firms of asking lawmakers to “legitimise larceny.” He also said they were trying to strong-arm Australian policymakers with the message: “if you don’t do this, we won’t invest.” The quote captures the core tension in the debate. Tech companies present copyright reform as a condition for investment and innovation. Rights holders see the same request as an attempt to rewrite property rules after the fact, making lawful what they regard as unauthorised use.
The investment argument is powerful in a country trying to build a local technology industry. Governments do not want to be left out of the AI economy, and they are wary of companies choosing to train models in jurisdictions with friendlier rules. But the ICMP’s response is that a race to the bottom on copyright would not create a healthy AI industry. It would transfer value from creators to platforms and make it harder for artists to earn from the very works that make AI outputs commercially attractive.
Australia’s copyright dilemma
Australia has become a focal point because it appears to be the lone country giving the industry’s request a second look. Tech companies have reportedly asked dozens of governments to relax copyright rules for AI, and most have either rejected the push or moved cautiously. Australia’s review has alarmed creative industries that have long suspected tech giants were quietly lobbying for looser rules behind the scenes. Those suspicions gained weight when lawmakers from both sides on a parliamentary AI committee brought the topic back up and described the situation as a deadlock.
The committee’s argument is that AI models in the United States already learn from copyrighted material without creators getting paid. Meanwhile, the companies building the most advanced AI models reportedly will not train them in Australia unless local copyright laws change. That framing puts Australia in an awkward position. If it refuses to relax the rules, it may lose some inward investment. If it gives in, it would be the odd one out among the countries tech firms approached, and it would expose Australian artists to a system in which their work feeds AI without a royalty check in sight.
Prime Minister Anthony Albanese took a firm public stance in July. He called it “theft” if creators could not decide whether AI learns from their work or how much they earn from it. That language matters because it frames the issue in terms of consent and compensation, not just technical copyright exceptions. The question is whether that stance will hold as industry pressure builds and as the government weighs promises of investment against the concerns of Australia’s creative sector.
What a weaker copyright regime could mean for artists
If Australia relaxes its rules, the practical effects could be subtle but far-reaching. AI developers might gain a broad exception for text and data mining, allowing them to ingest copyrighted music, books, articles and images without individual licences. Rights holders could be left with an opt-out system, which places the burden on artists to discover that their work has been used and then prevent future use. Opt-outs are often difficult to enforce, especially when datasets are large, opaque and assembled across borders.
A licensing system would work differently. It would require AI companies to negotiate with publishers, labels, collecting societies and artists, or to pay into a collective scheme. That could create a new revenue stream for creators and give them some control over how their work is used. Critics of licensing argue that it could be slow, costly and difficult to apply to models trained on billions of works. Supporters counter that copyright has always adapted to new technologies through licensing markets, and that AI should not be treated as an exception simply because the scale is large.
Transparency is another flashpoint. Artists and publishers cannot enforce rights if they do not know what has been ingested. Calls for dataset disclosure, record-keeping and attribution have grown alongside the legal debate. Without that information, creators are left to infer from model outputs whether their work was used. That is a weak basis for negotiation, especially when AI systems can generate outputs that are similar in style without copying a protected expression in a way that is easy to prove.
Why the dispute is bigger than one pop star
The Minogue and Cave claims sit inside a much larger global argument about AI and creative labour. Authors, visual artists, photographers, journalists and musicians have all raised objections to models trained on their work without permission. Some have sued. Some have licensed their archives. Others have called for new laws that require consent, compensation and transparency. The music industry is particularly exposed because songs are both cultural products and data-rich assets. Lyrics, melodies, rhythms and recordings can all be analysed, and the resulting models can be used to generate new music, imitate styles or provide production tools.
For Australian artists, the stakes are not only financial. A copyright regime that permits wholesale ingestion without payment could weaken the country’s cultural ecosystem. Songwriting, performance and production are risky careers that depend on copyright revenue. If AI companies can use Australian music to build commercial products while refusing to pay for it, the incentive to invest in local creative work may shrink. The artists who break through globally, as Minogue did, may still have leverage. The ones trying to follow in her footsteps may not.
The ICMP’s warning also challenges the idea that AI policy can be separated from cultural policy. Governments often treat copyright as a technical area for lawyers and trade negotiators. But copyright rules determine who gets paid, who controls distribution and who can build a business on the back of creative work. When those rules are rewritten for AI, the effects ripple through the music industry, the publishing industry and the wider creative economy.
For now, Australia’s position remains unsettled. The government has sent mixed signals, with strong rhetoric from the prime minister and a parliamentary committee describing deadlock. Tech companies continue to argue that investment depends on flexible rules. Publishers and artists continue to argue that flexibility cannot mean free use. Minogue’s name has become attached to the dispute because her catalogue is valuable, recognisable and commercially alive, but the principle at stake applies to every songwriter whose work has been swept into a dataset without a licence.
The committee’s next move will determine whether that warning becomes policy or remains a lobbying pressure point.
Source:Digital Trends News