The people who make Australian music have run out of patience with generative artificial intelligence. Songwriters, performers, publishers and the labels that back them are pressing the federal government to confront what they describe as wholesale copyright infringement, arguing that AI systems capable of producing convincing songs have been built on decades of recorded work that was never licensed and never paid for.
The frustration has been building for the better part of two years, but it has sharpened as text-to-music tools have improved to the point where an amateur can generate a passable track in seconds. Behind that convenience sits a harder question the industry wants answered: what did these models learn from, and did anyone consult the artists whose catalogues appear to have been swept up in the training data? As reported by MSN, the sector is now demanding a clear response rather than another round of polite consultation.
How the fight got here
Australia’s copyright framework predates generative AI by decades, and it does not deal neatly with the practice of feeding millions of protected works into a model so it can learn to imitate them. Unlike the United States, Australia has no broad “fair use” doctrine, only a narrower set of fair dealing exceptions, which means that copying a song for the purpose of training a commercial system sits in genuinely contested territory. Rights holders read the law as clearly requiring a licence. The technology companies argue that machine learning is a transformative process that does not reproduce the underlying works, so it should not trigger the same obligations.
That standoff has played out inside a broader review of how Australia should govern AI, and copyright has become one of its most divisive threads. The Albanese government has signalled it wants the country to capture the economic upside of AI while protecting creators, but those two goals have proven difficult to reconcile in practice. The tension surfaced publicly during the Prime Minister’s own remarks on the technology, which exposed a split within Labor over where the line between innovation and appropriation should fall.
What the industry is asking for
The music sector’s core demand is straightforward: models trained on Australian recordings and compositions should be licensed, and the creators behind that material should be paid. Beyond payment, the industry wants transparency obligations that would force AI developers to disclose what they trained on, so that infringement can actually be detected rather than merely suspected. Without disclosure, artists argue, they are being asked to prove a breach they cannot see inside a black box.
There is also a push for consent to sit at the front of the process. Peak bodies representing songwriters and rights holders have consistently argued that an “opt out” model, where an artist’s work is used unless they actively object, reverses the basic logic of copyright. Their preferred position is opt in, where nothing is used until permission is granted. The distinction sounds technical, but it decides who holds the power in every negotiation that follows.
The other side of the argument
Developers and some economists counter that a licensing regime strict enough to satisfy the music industry could kneecap Australia’s ability to build competitive AI at home. Their case is that training data is drawn from vast, global sources, that individual Australian tracks make up a vanishingly small share of any large model, and that requiring case-by-case licensing would be so expensive and slow that serious development would simply happen offshore, in jurisdictions with looser rules. On that view, a hard line on copyright would not protect Australian artists so much as export the industry that might otherwise employ them.
Consumer and open-technology advocates add a further point. They worry that a regime built to please large rights holders could entrench the biggest players on both sides, the major labels and the major AI firms, while smaller independent musicians and independent developers are left with the legal costs and none of the leverage. The debate, in other words, is not a simple contest between artists and machines. It runs through questions of scale, market power and who can afford to litigate.
Why this matters for Australia
Music is not a marginal part of the Australian economy, and the stakes here reach well beyond a single sector. The creative industries employ hundreds of thousands of people and feed a cultural export story that Australia has spent decades building, from pub rock to global pop. If the value that flows from recorded music can be captured by systems that pay nothing back to its creators, the incentive to keep making that music weakens. That is the outcome artists say they are trying to prevent.
There is also a sovereignty dimension that connects this fight to the wider national conversation about AI. Australia has been debating how much of its own AI capacity it should own and control, and copyright is part of that ledger. A country that wants a domestic AI industry still has to decide whether that industry is allowed to be built on the uncompensated labour of its own artists. Get the settings wrong in one direction and creators are hollowed out. Get them wrong in the other and the whole endeavour moves overseas, taking the jobs and the tax base with it.
What happens next
The immediate pressure is on the federal government to translate its stated support for creators into enforceable rules, whether through amendments to the Copyright Act, a dedicated AI transparency requirement, or a licensing scheme negotiated with the sector. The recently established national AI office gives Canberra a vehicle to coordinate that response, and artists have already argued they deserve a permanent seat at that table rather than an occasional invitation.
None of it will be quick. Copyright reform is slow, the technology is moving faster than any statute can track, and the government is trying to keep both creators and investors onside at once. What has changed is the tone. The music industry is no longer asking whether AI raises copyright problems. It is asking what the government intends to do about breaches it says are already happening, and it wants an answer measured in policy rather than sympathy.
Sources: MSN.
