'It's All Faintly Ridiculous': Pond Wade Through Dissatisfaction For Hope On 'Terrestrials'

Thousands Of Australian Artists' Songs Stolen In AI Data Heist

Artists such as AC/DC, Kylie Minogue, INXS, Cold Chisel, Tina Arena, and more are among those whose work has been uncovered in these collections of data.

Cold Chisel, Kylie Minogue & AC/DC
Cold Chisel, Kylie Minogue & AC/DC(Credit: Peter Dovgan; Edward Cooke; Andrew Briscoe)
More AC/DC AC/DC

A new investigation has confirmed that songs by Australian and New Zealand musicians are among the millions which have been stolen for use in AI datasets.

The claims come from The Atlantic reporter Alex Reisner, whose ongoing investigations into the world of AI training data have uncovered massive datasets of songs which are reportedly being shared within the community focused around AI development.

According to Reisner, one of the four datasets uncovered include 12 million songs, another has nine million, while two others include more than 100,000 songs.

Together, they feature music from artists as beloved and prominent as Taylor Swift, Nirvana, The Beatles, Bad Bunny, and more. If one was to listen to the entirety of the largest dataset, it would take 91 years to do so.

As part of the reporting, a search tool has been set up, allowing users to peruse these datasets for musicians (or authors, actors, and more) to discover whether their work has shown up (but not necessarily used for) AI training.

However, utilisation of the tool discovers that countless Australian and New Zealand artists have had their own compositions – millions of them – utilised as part of these datasets.

Artists such as AC/DC, Kylie Minogue, INXS, Cold Chisel, Tina Arena, and more are among those whose work has been uncovered in these collections of data.

"Midnight Oil. Sia. Crowded House. Lorde. Yothu Yindi,” says APRA AMCOS Chief Executive Dean Ormston. “This week, AI companies are asking the Australian and New Zealand Governments for a copyright carve-out.

“This week, we can show you exactly what they have already taken. No permission. No licence. No payment. These are not bargaining chips, they are the life's work of Australian and New Zealand songwriters.”

In October, the Federal Government ruled out a proposed text-and-data-mining exception in the Copyright Act following backlash from artists and music industry groups. Since then, other governments have followed Australia’s lead to do so, thereby protecting the rights of musicians and creators.

“Australian creatives are not only world-class, but they are also the lifeblood of Australian culture, and we must ensure the right legal protections are in place,” said Attorney General, Michelle Rowland at the time.

"Major tech platforms have not come to the table. Not once,” continued Ormston. “Instead they have lobbied governments, circulated policy papers, and proposed solutions designed to extinguish any obligation to pay.

“The only path forward is a genuine licensing conversation with the people whose work they have been using. We are ready. We have always been ready. The question is whether they are.”

According to APRA AMCOS’ landmark AI And Music Report, in the absence of a mandatory licensing framework, songwriters and composers from Australia and New Zealand face revenue losses of 23% – ultimately resulting in more than $500 million lost in just over four years.

Additionally, numerous Aboriginal and Torres Strait Islander musicians have also had their work stolen as part of scraping for datasets, including the likes of Gurrumul, Warumpi Band, William Barton, Christine Anu, Dan Sultan, Emma Donovan, and more.

Māori artists such as Stan Walker, Six60, Maisey Rika, Marlon Williams, and Horomona Horo have also had their music discovered in the datasets.

:Aboriginal and Torres Strait Islander artists have not given permission for their work to be used to train these AI datasets,” stated APRA AMCOS Director of Aboriginal and Torres Strait Islander Programs and Strategy, Leah Flanagan. “Many of these recordings carry cultural knowledge, language and connection to Country.

“Their use without consent is not acceptable, these are not just recordings, they are cultural expressions governed by protocols as well as copyright.

“In some cases, this is material that has never been available for commercial use by anyone, on any term,” Flanagan added. “Any response to AI and copyright must address Indigenous Cultural and Intellectual Property specifically and must be developed with First Nations communities at the centre.”

Just this month, CISAC (for which Ormston serves as Chair) held their General Assembly, with APRA AMCOS joining those who signed the Paris Commitment, which sets out a series of principles urging governments, technology companies, and cultural industries to ensure human creators remain protected, recognised, and fairly remunerated in the AI era.

“The Australian Government made the right call last October,” Ormston adds. “The pressure being applied right now by tech platforms alongside the Tech Council of Australia and Business Council of Australia is designed to reverse it.

“Our message is the same as it has always been: we are ready to license. We know how to do it. We have been doing it for over 100 years. But it has to be a real licensing framework, not a carve-out dressed up as a compromise.”

The Atlantic’s AI dataset search tool is available to use now.