Australia Bans AI-Generated Songs From Its Music Charts

Songs largely or wholly created by AI have been banned from Australia's music charts. From this week, all releases must be "substantially human made" to be eligible.

The Australian Recording Industry Association said the new code would help "promote the human nature of artistry."

The case that prompted it

The rule follows a controversy over a cover of Madonna's Like A Prayer by the Australian DJ Josh Fawaz, which topped the ARIA dance singles chart and reached number two on the overall chart.

The track became a staple of commercial radio playlists and has been streamed more than 48 million times on Spotify alone. Fawaz later added generative AI credits to the song following a backlash.

That sequence is the important part. The disclosure came after commercial success, not before it, and there was no mechanism that would have surfaced it earlier.

Songs must be 'substantially human made' to be included in Australia's charts under the new ARIA code.

ARIA's reasoning

The association said Australian artists are competing for attention "in the most crowded market in history" and that it was "not interested in promoting or celebrating the success of AI-generated music that does not contain human artistry."

That framing treats the chart as something other than a neutral measurement. A chart that simply counts streams reports what happened; a chart with an eligibility code makes a judgement about what deserves to be counted. ARIA has explicitly chosen the second, and the justification is that the chart functions as promotional infrastructure for an industry rather than as a statistic.

Where the line sits

Under the new rules AI assistance can still be used. The requirement is that the work be substantially human made, which preserves the enormous range of legitimate machine involvement in modern production — pitch correction, stem separation, mastering, arrangement tools, synthesis.

The difficulty is that "substantially" has no technical definition, and the boundary between a producer using generative tools and a producer prompting a generative system is genuinely blurred. Every practical enforcement question — who assesses it, on what evidence, and what happens to a track already charting — sits in that ambiguity.

The enforcement problem

The Fawaz case illustrates it precisely. The AI involvement became known because he disclosed it under pressure, not because anyone detected it.

Reliable detection of generative audio is not currently possible at scale, and the tools that claim to do it perform poorly on heavily processed commercial music, which is by definition already synthetic-sounding. That leaves the code dependent on self-declaration, backed by the threat of reputational damage if an undisclosed track is later exposed.

That may be enough. Charts have always run partly on industry norms rather than verification, and a rule that makes non-disclosure a career risk changes behaviour even when it cannot be policed directly.

Why charts still matter

It would be reasonable to ask whether a chart position means anything in a streaming era, and the answer is that it means more than it appears to.

Chart placement drives radio playlisting, playlist inclusion, festival booking and press coverage. It is one of the last remaining mechanisms by which an unknown artist becomes a known one, and the Fawaz track's journey onto commercial radio playlists shows the machinery working exactly as designed.

That is why ARIA is defending the eligibility rules rather than the measurement. A chart is a gate, and the association has decided what may pass through it.

What other markets will watch

Australia is a mid-sized music market that has moved first, which makes it a useful test.

If the code proves administrable — if disclosure holds, and disputes are rare — larger markets have a template. If it collapses into unresolvable arguments about how much assistance is too much, that will be equally informative, and rather more likely.