One of the central questions of the artificial-intelligence boom is deceptively simple: when an AI system learns from books, songs, articles and images, should the people who created that material be paid? In Australia, that question drew musicians and authors to Parliament House in Canberra this week, and it has put the government on a collision course with some of the world's biggest technology companies.
Creators take their case to Parliament
Australian artists and industry bodies gathered at Parliament House with a blunt demand: that AI firms seek permission and pay before using copyrighted work to train their models. The rally brought together the music and publishing sectors, and organizations representing songwriters and performers framed the mass use of their catalogues to build AI systems in stark terms. The head of the songwriters' and composers' body APRA AMCOS, Dean Ormston, called it "the largest theft of intellectual property in the history of our industry," Rolling Stone Australia reported.
The anger has a concrete trigger. Reports have indicated that large volumes of Australian music — including work by well-known artists — were swept into datasets used to train AI, without the creators' knowledge or consent. For songwriters, authors and performers, the concern is not abstract: AI systems trained on their work can generate material that competes with it, and they say they should be compensated for supplying the raw ingredients.
What the government has decided
The creators are, for now, pushing on a partly open door. Australia's government has already rejected the most sweeping proposal to loosen copyright for AI. That idea came from the Productivity Commission, an independent advisory body, which floated adding a "text and data mining" exception to the Copyright Act — a change that would have let AI developers train on copyrighted works without permission or payment, the Kluwer Copyright Blog reported.
The government declined to adopt it, according to The National Law Review. The arts minister, Tony Burke, said the government had "no plans, no intention, no appetite" to weaken copyright law on the basis of the proposal, signaling a preference instead for a licensing approach in which AI firms negotiate and pay for the content they use. That leaves the onus on the technology companies rather than on the creators.
The case on each side
The argument is genuinely contested. Technology companies — among them Google, Meta and OpenAI — and some economists argue that training AI on existing material is more akin to learning than copying, that licensing every work individually is impractical at the scale involved, and that overly strict rules could deter investment and leave a country behind in a strategically vital field. The Productivity Commission's report pointed to large potential economic gains from AI.
Creators counter that their work has real value, that letting profitable firms build products on it without payment is simply unfair, and that their livelihoods are at stake as AI tools spread. Notably, the principle that AI firms should pay has drawn support across Australia's political divide, with figures in the opposition also backing the creators' position even amid wider partisan disagreement. Public opinion appears to favor a middle path: survey work cited in the debate has suggested most Australians would accept AI training on copyrighted material provided creators are fairly compensated.
A global fight
Australia's stance is one front in a battle unfolding worldwide. In the United States, a wave of lawsuits pits authors, artists, news organizations and music companies against AI developers over whether training on copyrighted work is permissible "fair use"; one prominent case saw the AI company Anthropic agree to a large settlement with authors, and further suits are proceeding. The European Union's rules require AI providers to respect rights-holders' ability to opt out of having their work mined and to be more transparent about training data. In Britain, a plan to make training on copyrighted material easier met fierce resistance from the creative sector, and a UK court recently grappled with whether an AI model's internal data amounts to a "copy" of the works it learned from.
The common thread is a search for balance between two things societies say they value: a thriving AI industry and a thriving creative one. Different governments are drawing the line in different places.
Why it matters
Decisions like Australia's are early attempts to write the rules of an economy in which human-made content is the fuel for machine-made output. Requiring payment pushes AI companies toward licensing markets that could channel money back to creators; allowing free use lowers costs for developers but risks hollowing out the industries the material comes from.
Australia has signaled which way it leans, and its artists are pressing it to hold that line. But with courts, parliaments and regulators around the world still working through the same question, the terms on which AI and human creativity coexist remain unsettled — and the answers reached in the next few years are likely to matter for a long time.
