A narrower rule than the headlines suggest
The Australian Recording Industry Association (ARIA) announced on August 25, 2026 that it has updated its Charts Code of Practice to exclude wholly AI-generated recordings from the ARIA Charts and ARIA Awards, effective with the chart dated August 31 and published August 28 (ARIA). Read the coverage quickly and it sounds like a blanket ban on AI in music. It isn't. ARIA's own language draws a narrower line: recordings that use generative AI in a supporting role remain eligible provided they are "substantially human-made" and raise no concerns about stream or chart manipulation. What's excluded is music where AI "creates the entirety or primary portion of the creative elements" — the distinction ARIA borrows from a global industry labelling standard published in July.
The trigger was a chart-topping AI cover
The timing is not abstract policy-making. Gold Coast producer Josh Fawaz's AI-assisted cover of Madonna's "Like a Prayer" became Australia's most-played radio track and reached number four on two ARIA charts in July, with Spotify crediting generative AI for the drums and vocals (SBS News; ABC News). That a synthetic vocal performance could out-chart human artists using licensed samples of a decades-old song is precisely the scenario ARIA's code now forecloses. ARIA CEO Annabelle Herd framed the stakes bluntly: a chart that rewards unlicensed AI output would "undercut the very basis of the recorded music we exist to present," while adding that the new rule "gives everyone who releases music in Australia a clear rule to work to."
Steelmanning the ban
The case for ARIA's move is stronger than "industry protectionism." Charts are not neutral popularity trackers — they are curated products that labels, radio programmers and award shows treat as a proxy for genuine audience response to human artistry, and ARIA is a private body entitled to define what its own product measures. More importantly, the AI systems capable of generating a fully convincing vocal and instrumental track were, in nearly every commercially available case as of 2026, trained on copyrighted recordings without a license from the rights holders whose stylistic and sonic fingerprints they reproduce. A chart position translates into streaming placement, sync licensing interest and touring leverage; letting output from unlicensed training compete for that position with human artists who paid session musicians and cleared samples is a real market distortion, not a hypothetical one. Electronic artist Adam Hyde of Peking Duk put the underlying grievance plainly: "AI is using our stuff. It's using a lot of our peers' stuff."
Where the rule gets it right
The strongest feature of ARIA's approach is exactly what makes it defensible from an innovation standpoint: it does not ban AI tools, it bans a specific outcome — wholly synthetic output competing as though it were human performance — while leaving AI-assisted production, mixing, mastering and songwriting aids untouched. That mirrors the framework the International Federation of the Phonographic Industry (IFPI) rolled out on July 30, 2026, which conditions eligibility on the AI service being lawfully licensed, the recording being substantially human-made, and the release being free of stream-manipulation concerns — not on AI's mere presence in the production chain (IFPI). IFPI is applying the principles directly across 13 chart markets in Latin America, the Middle East, Africa and Southeast Asia, and coordinating adoption through more than 20 national chart bodies including ARIA — a genuinely global, industry-led standard rather than one country improvising alone.
The unresolved question is licensing, not labelling
Where the policy is incomplete is that it treats a symptom — chart placement — without touching the underlying dispute: whether AI developers need a license to train on recorded music at all. That fight is still unresolved in Canberra. The Attorney-General's Copyright and Artificial Intelligence Reference Group (CAIRG), established in December 2023, spent late 2025 examining licensing arrangements and legal clarity for AI-generated material, and as of the government's April 2026 IP report, no legislative text-and-data-mining exception has been adopted and no licensing framework has been finalized (IP Australia). ARIA's chart rule is, in effect, the music industry building a private enforcement mechanism because the public one is still years from draft legislation.
A proportionate model, if it stays narrow
For a pro-innovation, evidence-based line, ARIA's rule is closer to right than wrong: it targets unlicensed displacement of human creative labor rather than the technology itself, it preserves a large and growing category of AI-assisted work, and it builds in a dispute process for artists to contest exclusion with evidence. The risk is scope creep — if "substantially human-made" is applied inconsistently, or if the definition drifts toward penalizing any AI involvement rather than unlicensed training, a sensible anti-fraud measure could calcify into a de facto tax on legitimate AI-assisted Australian artists. With draft copyright legislation not expected before 2027, ARIA's code will be doing the real regulatory work in this market for at least another year — which is reason enough for the government to move the licensing question off the reference-group track and onto a statutory one.