Artists report their catalogs being scraped, threatening licensing terms and nullifying agreements already in place.

Artists warn that AI training using their music erodes control over licensing and compensation, insisting contracts mean nothing if usage is unlicensed.

Dempsey argues that AI can reproduce his work without paying, undermining negotiated terms; Fanning warns AI dehumanizes art by machines aggregating human emotion.

Dempsey notes his Something For Kate and solo works appear in the datasets, describing it as a violation that undermines contractual terms.

The Atlantic identifies two large datasets, Sleeping-DISCO-9M and LAION-DISCO-12M, that include Australian songs linked to AI training concerns.

Australian musicians, including Dempsey, Fanning, and Hayes, join protests over AI training on their originals without permission.

The Atlantic notes that inclusion in datasets does not prove actual use in AI training and may omit some uses by AI companies.

APRA AMCOS, representing 128,000 members, calls the datasets evidence of creative theft and faults major tech platforms for non-engagement and lack of fair licensing.

APRA AMCOS frames the datasets as theft and criticizes platforms for failing to engage creators or compensate for use.

APRA AMCOS condemns permissionless use of members’ works and accuses platforms of lobbying against fair compensation.

Australia’s copyright regime generally requires permission and negotiated terms for use, with policy moves on text-and-data mining stalled after a Productivity Commission review and government reversal in late 2025.

The policy debate around AI content use in Australia has featured the Productivity Commission, but proposals to loosen pay requirements were ruled out by the government in October 2025.