The Australian music industry is opposed to the Productivity Committee's proposal to introduce new fair trade exceptions in Text and Data Mining (TDM) in AI Training. Essentially, it legalizes AI to be trained with copyrighted materials, including music.
Adam Moses, the creative head of mosaic music and sound, tells LBB when AI training gives blanket copyright exceptions, threatens “a dark world where it is impossible to say who actually owns the music as a result” and “Blurs the author.”
“When everything is trained from everything, the very concept of ownership becomes irrelevant, which undermines the very foundation of our industry.
The Australian Recording Industry Association (ARIA) and the Australian Gramophone Performance Company (PPCA) claim that the move will destroy copyright protections, reduce creative work and strip away potential license revenues from local artists.
“We're looking forward to seeing you in the future,” said Annabelle Herd, CEO of Aria and PPCA. “This is an existential moment. Protect Australia's right to stay creativity or watch the cultural outcomes fall apart by global tech companies while creators lose the ability to make a living from their jobs.”
If the exception is passed, it could “wipe away the $500-$600 million annual AI licensing market for Australian creators and cause a permanent market collapse in the sector employing more than 1 million Australians,” she said. “It would legalize theft of the cultures of Aboriginal and Torres Strait Islanders and destroy cultural sovereignty.”
This week, Australia's indigenous rapper Briggs Appears in the Senatejoins an Australian creative panel who speaks against the proposed exception.
When independent Senator David Pocock asked the musician if AI had the ability to forge Brigstruck in its current form, he said, “I'm very suspicious of that. I don't think AI understands Shepperton's lounge room at this time.
The report argues that exceptions can drive innovation, but Annabelle opposed.
“From Merlin and Kobalt licenses to ElevenLab licenses, to Google-licensed AAP content in collaboration with News Corp, hundreds of AI licensing transactions already operating at scale prove that the current framework is working, before we acknowledge the fact that digital companies like Netflix, Spotify and YouTube are based on licenses for the content.”
Instead, she argued that the real barrier to AI growth is access to capital and infrastructure, not copyright. “The exception to TDM would strip Australian creators and rights holders of property rights to maintain the industry that donates 6-7% of GDP. [gross domestic product] It employs over 1 million Australians. ”
The proposal to go ahead will change the way he negotiates transactions with his brand, agents or other clients, Adam added.
“If AI companies can use recordings for free, there is a risk of a more broadly assessing music rights, which makes it difficult to assert fair rates with brands and agents who may begin to view music as cheap or unlimited resources.
Potential risks include expanding license revenue, eroding copyright respect, and allowing AI-generated tracks to “compete with human music without the investment, risk, or originality needed to create it.”
Adam argued that licensing is the solution and that AI platforms “pay for music to train the system, as does brands, agents or broadcasters.” He added that the royalty model is fair and reflects the sample's license that “credit and payment remain linked to the original work.”
Adam also wants transparency. “If my music is used to train AI, I need to know when, where and how I do it.”
Annabelle added that the industry should advocate for ethical partnerships, push back on proposals that allow for unauthorized use, and assert transparency around AI training data. In the US, where there is no exemption from text and data mining, “there are currently more than 50 cases to violate the limits of fair use,” Annabelle said.
“Strong copyright and strong AI investments have been successful. Private AI investments in the US reached US$100.1 billion in 2024, reaching 12 times the number of China and 24 times the number of UK without the TDM exception.
“Nationals with TDM exceptions have not seen groundbreaking investments. The EU has acknowledged that the 2019 administration is suited to generative AI. The lesson is clear: governments must stay in the way to maintain strong copyright and make the free market work.”
