🇦🇺 AUSTRALIA “BANS” AI MUSIC? NOT QUITE — AND THAT IS WHERE IT GETS INTERESTING
THE HEADLINE IS SIMPLE. THE REALITY IS NOT.
AI • Human Creativity • Music Charts • Government Policy • Technology • Labour • Streaming Economics • Copyright • Institutional Power
By Gordon Lawrence Taylor | Infectious Unease Radio – Subterranean Zone Radio – Voice In The Void | Research current to 31 August 2026
Only days after two substantial Voice In The Void investigations examined artificial intelligence, computer music, human labour and the boundaries of creative authorship, another AI story emerged from Australia carrying an irresistible headline: AUSTRALIA BANS FULLY AI-MADE SONGS FROM OFFICIAL MUSIC CHARTS.
It sounds decisive. Almost revolutionary. Human musicians, apparently, have been protected from the approaching machine.
Except that this is not quite what happened.
Australia has not banned AI music. The Australian Government has not made AI-generated recordings illegal, has not prohibited their release and has not ordered streaming services to remove them. The Australian Recording Industry Association — ARIA — has changed the eligibility rules for the ARIA Charts. From the chart dated 31 August 2026, wholly AI-generated recordings are ineligible, while recordings using generative AI in a supporting role remain eligible where they are substantially human-made and satisfy the other chart requirements.
That distinction transforms the argument. This is not really a story about Australia banning artificial intelligence. It is a story about an industry attempting to decide how much meaningful human participation must remain inside a recording before that recording continues to qualify as human musical expression.
And that is considerably more interesting.
The immediate spark for this follow-up came from the wider discussion around a post circulated by AI Councillor, arriving almost immediately after the same questions had already been explored through Voice In The Void. AI Councillor
THE SONG THAT FORCED THE QUESTION
The immediate Australian controversy centred on producer and DJ Josh Fawaz and his version of Madonna’s Like a Prayer. The recording became a major radio success and appeared across ARIA charts. After scrutiny surrounding the production, Spotify’s credits were updated to identify generative-AI vocals and AI drums. ABC reported in July that the recording had become number one on Australia’s national airplay charts and had already generated tens of millions of Spotify streams.
That creates an awkward question. If a producer selects the concept, directs the technology, structures the arrangement, chooses between outputs, edits the result, mixes the recording and releases it — but the principal singing voice was artificially generated — who actually performed the record?
The producer unquestionably contributed creative labour. But who sang it? And just as importantly, which human performer was no longer required?
That final question moves the story beyond novelty and directly into labour, authorship, copyright and economics.
ARIA HAS NOT BANNED AI — IT HAS ATTEMPTED TO DEFINE ITS ROLE
ARIA’s actual policy is considerably more nuanced than the headline. An AI-generated lead vocal makes a recording ineligible. So can a key AI-generated instrumental performance. A recording generated entirely from prompting is similarly classified as AI-generated.
But human lead vocals with AI-generated backing vocals can remain eligible. AI mastering, stem separation, reverb, instrument patches, short background samples and specified AI production tools can also remain acceptable. ARIA even acknowledges that a human performance materially transformed through certain AI processes may still qualify as AI-assisted rather than AI-generated.
So artificial intelligence itself is not prohibited.
ARIA is attempting to distinguish between the machine as a tool and the machine as the principal performer or creator.
That sounds relatively simple until somebody attempts to measure it.
A synthetic backing voice might be acceptable. Put essentially the same synthetic voice at the centre of the recording and its status changes. The technology may not have changed. Its position within the creative hierarchy has.
There is no instrument in a studio displaying:
HUMAN CREATIVE CONTRIBUTION: 72% — MACHINE CONTRIBUTION: 28%.
Creativity cannot be measured like volume, frequency or voltage.
And that is precisely where the problem becomes intellectually interesting.
MACHINES WERE IN THE STUDIO LONG BEFORE GENERATIVE AI
The idea that machines have suddenly invaded music is historically absurd. Electronic and recorded music have been negotiating relationships between humans and machines for generations.
Synthesizers generate sounds that may have no acoustic equivalent. Drum machines mechanise rhythm. Sequencers execute musical events. Samplers turn previously recorded sound into new compositional material. MIDI separates musical instruction from the instrument producing the sound. Digital audio workstations replaced rooms filled with physical editing and recording equipment. Pitch correction alters vocal performance. Software instruments imitate acoustic instruments or create entirely synthetic ones.
Computer-assisted composition is hardly new either. David Cope’s work on Experiments in Musical Intelligence began in the 1980s, and his 1996 book documented computational attempts to understand and generate musical style.
Entire musical traditions are inseparable from technological intervention: techno, industrial, EBM, jungle, drum and bass, house, dub, hip-hop, ambient and experimental electronic music among them.
ARIA itself has repeatedly adapted its commercial measurements to technological change. Streaming was incorporated into the Singles Chart in November 2014 and into the Albums Chart in May 2017 because, in ARIA’s own words, the charts needed to reflect how audiences consumed music. By 2024, digital formats accounted for 91.5 per cent of Australian recorded-music revenue, with subscription services alone accounting for 71 per cent.
The meaningful question therefore cannot simply be:
WAS TECHNOLOGY USED?
It has been used for decades.
The better question is:
WHAT WAS THE TECHNOLOGY ASKED TO DO?
A conventional sequencer generally executes instructions supplied by the musician. A modern generative system can increasingly determine melody, harmony, instrumentation, arrangement, vocal characteristics and production from comparatively broad instructions.
Older technologies frequently ask:
What does the human want the machine to perform?
Generative AI increasingly makes possible a different question:
What should the machine generate for the human to choose from?
That is a substantial creative shift.
THE MACHINE IS NOT AUTOMATICALLY THE CREATOR
Using AI does not automatically remove human creativity.
A writer using sophisticated software remains a writer. A photographer using digital processing remains a photographer. A producer using electronic instruments remains a producer. A musician using generative technology does not automatically cease being a musician.
AI can repair recordings, separate stems, analyse material, suggest alternatives, reorganise information, create prototypes and generate fragments that a human subsequently rejects, selects, alters and reconstructs.
The distinction is also emerging in copyright law. The US Copyright Office concluded in 2025 that AI-assisted work can retain copyright where a human determines sufficient expressive elements; the inclusion of AI-generated material does not itself disqualify a larger human-authored work. Mere prompting, however, does not necessarily establish authorship.
That is very close to the underlying creative argument:
human direction matters.
But generative technology makes that principle vastly more complicated because the machine can now supply material that resembles autonomous creative performance.
WHAT WOULD WILLIAM S. BURROUGHS HAVE MADE OF ALL THIS?
William S. Burroughs died long before generative AI entered everyday cultural production, so attributing a definite opinion to him would be dishonest. But his work offers an unusually useful framework through which to examine the argument.
Burroughs and Brion Gysin deliberately destabilised traditional ideas about authorship. The cut-up method fragmented existing texts and rearranged them into new structures. Tape experiments cut, repeated, layered and disrupted recorded language. Material originating elsewhere could become something new through selection, juxtaposition and intervention.
Scissors did not become the author.
The tape machine did not become the writer.
The technology altered the creative method.
Burroughs and Gysin’s idea of the Third Mind becomes particularly provocative in the AI age: something emerging through a combination that cannot be reduced completely to either contributor.
A contemporary producer may supply intention, constraints and judgement. A generative system supplies possible material. The human selects, rejects, changes and rearranges those possibilities. The final object can occupy an uncomfortable territory between traditional human authorship and autonomous generation.
Burroughs might have enjoyed that disturbance immensely.
But there is another side of Burroughs that matters far more to the present argument.
BURROUGHS WAS OBSESSED WITH CONTROL
Burroughs repeatedly returned to systems of control — language, media, bureaucracy, institutions and technologies capable of shaping behaviour and perception.
Seen through that lens, the most Burroughsian question about artificial intelligence may not be:
CAN THE MACHINE CREATE?
It may be:
WHO CONTROLS THE MACHINE?
Who owns the system? Who supplies its training material? Who determines what can be generated? Who owns the output? Who controls the algorithm deciding which recording becomes visible? Who receives the revenue? Who becomes economically unnecessary?
Those questions lead directly away from the machine itself and toward the institutions surrounding it.
BURROUGHS, ARIA AND THE THEATRE OF INSTITUTIONAL HYPOCRISY
From a Burroughs-like perspective, the present spectacle could look almost like institutional satire.
The Australian Government speaks about protecting human creators while simultaneously encouraging widespread AI adoption throughout the economy.
ARIA draws boundaries around synthetic creativity while representing an industry that has repeatedly embraced technological disruption when it expanded distribution, marketing and revenue.
Major music companies object to unauthorised generative AI while simultaneously negotiating licensed AI products.
Streaming companies warn about synthetic spam while developing artificial-intelligence products of their own.
The contradiction does not require a conspiracy. It requires only observing how large institutions frequently change their language depending upon context.
The machine is dangerous when it is uncontrolled.
The machine becomes innovation when it can be regulated.
The machine threatens artists when it trains without permission.
The machine becomes a revenue opportunity when licensing agreements are signed.
Automation threatens labour in one sentence and becomes productivity in the next.
From this perspective, Burroughs might have regarded the entire affair as a hypocritical bureaucratic joke: authorities attempting to regulate a machine they have already invited through the front door.
ARIA’s position contains an especially strange irony. An industry built upon recording technology, synthesis, sampling, digital editing, compression, algorithms, streaming and global data infrastructure is now attempting to determine precisely when technological intervention becomes too technological.
ARIA itself runs ARIA Innovator, an initiative explicitly concerned with rapid technological change, data-driven marketing, innovation and AI. Its March 2026 conference brought music and technology executives together to discuss AI, licensing, partnership and copyright.
That does not invalidate ARIA’s AI rule.
It does show that the organisation is not anti-technology.
The argument is about which technology, under which conditions, controlled by whom.
The machine is not necessarily the joke.
THE SYSTEM TRYING TO OWN THE MACHINE MAY BE.
AND THEN THERE IS ARIA ITSELF
Before treating ARIA’s decision as a neutral philosophical judgement about art, it is worth establishing what ARIA actually is.
ARIA is not Parliament. It is not an independent statutory tribunal determining the meaning of culture. It is an industry association embedded in Australia’s commercial recorded-music sector.
Its current Board of Directors includes representatives associated with ABC, Warner, Mushroom, Sony, Universal and Rubber Records. Its Copyright Committee comprises representatives from Warner, Universal and Sony, while the same multinational companies are strongly represented across its Finance and Chart & Marketing committees.
That structure does not establish corruption. An unsupported corruption accusation would weaken the documented argument.
The much stronger point is structural:
ARIA REPRESENTS THE RECORDING INDUSTRY WHILE ALSO ADMINISTERING IMPORTANT PARTS OF THE INDUSTRY’S CHART SYSTEM.
That makes scrutiny completely legitimate.
Industry self-regulation is not automatically improper. Charts require eligibility rules and somebody must administer them. But an organisation exercising commercial and cultural influence should not be confused with a detached philosophical authority deciding what music is.
ARIA IS ALSO FOLLOWING AN INTERNATIONAL INDUSTRY FRAMEWORK
The new Australian policy was not created in isolation.
ARIA explicitly states that its rules implement principles announced by the International Federation of the Phonographic Industry — IFPI — on 30 July 2026. IFPI’s framework is intended to be rolled out across official chart systems internationally and establishes requirements concerning human creativity, lawful and authorised AI services, manipulation, copyright and labelling.
That makes triumphant claims that Australia has independently invented a world-leading solution rather less persuasive.
A more accurate description is:
Australia’s recording-industry association has implemented a developing international recording-industry standard.
Less dramatic.
More useful.
And the licensing element exposes another fault line.
WHEN DOES AN UNACCEPTABLE MACHINE BECOME AN ACCEPTABLE LICENSED MACHINE?
Record companies have legitimate reasons to object when copyrighted recordings are used without authorisation to build commercial AI systems.
Creators deserve control, attribution and economically meaningful remuneration where their protected work is exploited.
But major music companies are simultaneously constructing authorised AI business models.
In October 2025, Spotify announced collaboration with Sony Music Group, Universal Music Group, Warner Music Group, Merlin and Believe on what it calls artist-first AI music products. In May 2026, Spotify and Universal announced licensing agreements for a paid generative-AI tool allowing users to make covers and remixes from participating catalogues, with consent, credit and compensation incorporated into the model.
This changes the dividing line from something morally simple such as:
HUMAN MUSIC versus AI MUSIC
into something commercially more revealing:
UNAUTHORISED AI versus AUTHORISED AND MONETISABLE AI.
That distinction makes sense in copyright law.
It does not answer the philosophical question about human performance.
A licence can address permission.
A royalty can address compensation.
A label can address disclosure.
But a licensing agreement does not transform a synthetic vocalist into a human vocalist.
The debates about copyright, labour, authorship and commerce are related — but they are not identical.
AUSTRALIAN AI POLICY — LEADERSHIP OR CATCH-UP?
Australia has become increasingly active around artificial intelligence, but technological leadership should not be declared simply because a new announcement has been made.
The European Union’s AI Act entered into force on 1 August 2024, establishing a binding, risk-based legal framework after years of legislative development. Australia, meanwhile, spent 2024 consulting on proposed mandatory guardrails for high-risk AI before deciding not to proceed with that particular model at that time. Its broader National AI Plan did not arrive until 2 December 2025.
That does not mean Australia has done nothing.
It means the Australian model has been comparatively incremental and reactive: consultation, voluntary guidance, adaptation of existing law, policy development and observation of international approaches before firmer interventions emerge.
There are rational arguments for caution. Technology develops faster than legislation, and badly written rules can become obsolete almost immediately.
But caution and leadership are not synonyms.
Following developments elsewhere, establishing another consultation and eventually adopting a framework is not necessarily evidence of being at the technological frontier.
THE AUSTRALIAN GOVERNMENT IS NOT ANTI-AI
Any suggestion that the Australian Government is philosophically opposed to artificial intelligence collapses under its own policy documents.
The National AI Plan explicitly aims to build an AI-enabled economy that is more competitive, productive and resilient. It seeks greater investment, broader business adoption, expanded skills and more AI-supported public services.
The government committed $17 million to AI Adopt Centres designed to help small and medium businesses implement artificial intelligence, with individual grants ranging from $3 million to $5 million.
On 1 April 2026, it signed an AI collaboration memorandum with Anthropic. On 23 April it signed another with Microsoft, welcoming Microsoft’s proposed $25 billion investment in Australia’s AI-enabled economy and a plan to train three million workers.
On 28 August 2026 — only days after ARIA’s announcement — the government backed the Buy Australian AI Partnership, connecting domestic AI businesses with major corporate buyers representing more than $22 billion in annual procurement spending.
This is not anti-AI policy.
IT IS ACTIVE AI INDUSTRIAL POLICY.
PROTECT THE CREATOR — AUTOMATE THE OFFICE?
The real contradiction lies in the language surrounding different kinds of labour.
When artificial intelligence potentially displaces musicians, discussion turns toward human creativity, copyright and cultural protection.
When artificial intelligence enters a corporate office, the vocabulary changes to efficiency, innovation and productivity.
Those benefits can be genuine. If software removes hours of repetitive work and gives a person more time for more valuable activity, AI can improve working life.
But a different possibility exists.
If six employees plus software can produce what previously required ten employees, the meaning of productivity changes considerably for the four workers whose labour is no longer required.
Government policy frequently assumes that AI will augment workers, create new industries and improve job quality. Those outcomes may occur. But governments do not make every commercial employment decision.
Employers do.
That makes a very simple economic question unavoidable:
WHO RECEIVES THE PRODUCTIVITY BENEFIT — AND WHO ABSORBS THE DISPLACEMENT?
GOVERNMENT ITSELF IS USING AI
Artificial intelligence is already entering the Australian Public Service.
GovAI Chat is a government-managed environment allowing APS staff to work with commercial models including ChatGPT and Claude. The Department of Finance describes it as an experiment in how generative AI can safely support public-service work.
Again, there is nothing inherently objectionable about this.
It actually reinforces the central argument.
AI is a tool.
The ethical significance depends upon what it is being used to do, what human responsibility remains and what consequences follow.
GOVERNMENT AI ALSO ENTERS MARKETING AND CREATIVE COMMUNICATION
The contradiction becomes particularly visible when artificial intelligence moves into creative government work.
The National AI Centre’s own transparency framework provides examples of AI-generated content including poster artwork produced from a verbal prompt and an AI-generated social-media post reviewed by staff.
The eSafety Commissioner states that its AI uses can include image processing and generation, AI-generated voice and support for design and education/marketing campaigns, while retaining human review and accountability.
Those uses are not automatically unethical.
But if replacing a singer raises concern about the human singer who might otherwise have been employed, the same question should not disappear when the machine generates artwork or marketing material.
What about the illustrator?
The graphic designer?
The photographer?
The copywriter?
The voice actor?
The communications contractor?
Human creative labour does not become less worthy of economic consideration merely because it happens outside a recording studio.
THE JOB QUESTION IS ALREADY MORE THAN THEORY
There is an important qualification. Evidence does not currently establish mass Australian unemployment caused by generative AI.
Jobs and Skills Australia reports that large-scale job displacement has not yet emerged and that many early effects involve altered tasks, upskilling and redeployment. But the same research finds reduced demand for some routine and clerical work and specifically identifies voice-over artists as a task-specific occupation already experiencing reduced demand.
Its case studies go further. Participants described entry-level work in voice-over, junior design and freelance content creation being displaced because generative tools can produce acceptable material rapidly and at little or no cost. One organisation told the researchers that AI output was already replacing substantial amounts of voice-actor work.
That matters enormously in an article about artificial singers.
A synthetic voice looks like an impressive technical achievement from inside the software laboratory.
It looks rather different from the perspective of the human performer no longer receiving the booking.
THE PRODUCTIVITY DOUBLE STANDARD
A strange institutional vocabulary begins to emerge.
AI threatening copyright becomes a protection issue. AI generating the lead vocal becomes a chart-integrity issue. AI inside government becomes responsible adoption. AI in a business becomes productivity. AI attracting multinational investment becomes economic opportunity. AI reducing the requirement for certain human tasks becomes a transition and retraining problem.
Every individual position can be rationally defended.
Placed together, however, they create the appearance of a profound double standard.
The technology is treated as morally threatening where an institution does not control the consequences, yet increasingly attractive where that institution can deploy, license, regulate or profit from it.
That is the hypocrisy worth examining.
THEN THERE IS SPOTIFY
Spotify provides perhaps the clearest example of how complicated the modern AI morality has become.
The company has taken action against AI impersonation, synthetic spam, deceptive content and fraudulent streaming.
At the same time, Spotify is developing generative-AI products with some of the world’s largest music-rights companies.
That does not automatically indicate wrongdoing.
It demonstrates something more revealing:
the music industry is not fundamentally opposed to AI.
The real questions increasingly concern ownership, authorisation, licensing, transparency and remuneration.
SPOTIFY PAYS BILLIONS — THAT DOES NOT MEAN EVERY ARTIST PROSPERS
Spotify has a substantial factual answer to criticism of its economics.
The company says it paid more than US$11 billion to the music industry in 2025, taking lifetime payouts above US$70 billion. It also says independent artists and labels accounted for around half of 2025 royalties.
Those figures matter.
Ignoring them would be intellectually dishonest.
But they answer one question:
How much money entered the music-rights economy?
They do not necessarily answer:
HOW MUCH DID THE INDIVIDUAL MUSICIAN PERSONALLY RECEIVE?
Spotify pays rights holders. Depending upon ownership and contracts, those rights holders can include labels, distributors, publishers and other organisations. Spotify itself does not have visibility into what an individual artist ultimately retains after contractual arrangements further along that chain.
This distinction deserves emphasis:
ROYALTIES GENERATED ARE NOT THE SAME THING AS CREATOR REMUNERATION RECEIVED.
THE LOWER-STREAMING ARTIST FACES A DIFFERENT ECONOMY
Since April 2024, a track must achieve at least 1,000 Spotify streams during the preceding 12 months to enter Spotify’s recorded-music royalty-pool calculation. Spotify states that the total pool is not reduced; instead, money previously associated with very small allocations is redistributed across tracks that meet the threshold.
Spotify argues that these very small allocations frequently failed to reach artists anyway because distributor withdrawal thresholds could exceed the amounts generated. It also says the policy discourages attempts to manipulate streaming by uploading enormous quantities of low-stream tracks.
There is logic behind that explanation.
There is also an unavoidable consequence.
An independent artist can compose a song, perform every instrument, sing every lyric, finance the recording, pay for mixing, mastering, artwork and distribution, promote the release and attract hundreds of genuine listeners.
At 200 streams, the recording has not qualified.
At 500, it has not qualified.
At 750, it has not qualified.
At 999 Spotify streams in the relevant period, it still has not entered the recorded-music royalty-pool calculation.
Every creative element could be entirely human.
The recording is nevertheless below Spotify’s monetisation threshold.
That is rather difficult to ignore when the music economy suddenly speaks passionately about protecting human artists from machines.
ENORMOUS INDUSTRY REVENUE CAN COEXIST WITH MINIMAL CREATOR REMUNERATION
Spotify’s 2025 data shows striking growth at the upper end of its system. According to its Loud & Clear figures, 303,200 artists generated more than US$1,000; 81,100 generated more than US$10,000; 13,800 generated more than US$100,000; 1,540 generated more than US$1 million; and 80 generated more than US$10 million.
Those achievements are genuine. Streaming has unquestionably given many artists opportunities unavailable in previous periods of the recording industry.
But another Spotify statistic is equally revealing.
The 100,000th-highest-earning artist generated a little more than US$7,300 during 2025. Spotify presents this positively because the equivalent position generated only around US$350 a decade earlier.
That is substantial growth.
But consider the number from the opposite direction:
the 100,000th-highest-earning artist on one of the largest music services on Earth generated approximately US$7,300 across an entire year.
And once again, that is revenue generated on Spotify before whatever contractual arrangements apply downstream.
This is not evidence that Spotify pays nobody.
It demonstrates something much more important:
AN ENORMOUSLY PROFITABLE DIGITAL MUSIC ECONOMY CAN GENERATE SUBSTANTIAL AGGREGATE REVENUE WHILE LARGE NUMBERS OF INDIVIDUAL CREATORS RECEIVE ECONOMICALLY MARGINAL RETURNS.
MUSICIANS WERE FINANCIALLY VULNERABLE BEFORE GENERATIVE AI ARRIVED
UK government-commissioned research provides sobering context. Music Creators’ Earnings in the Digital Era found that 37 per cent of surveyed musicians earned £5,000 or less from music in 2019, 47 per cent earned less than £10,000, and 62 per cent earned £20,000 or less. Many creators relied upon teaching, live performance or other employment rather than recorded music alone.
Generative AI did not create that inequality.
Which raises an uncomfortable question:
WHERE WAS THE SAME URGENT DEFENCE OF HUMAN CREATIVITY WHILE HUMAN MUSICIANS WERE ALREADY RECEIVING DISPROPORTIONATELY LOW REMUNERATION FROM A HIGHLY PROFITABLE DIGITAL MUSIC ECONOMY?
AI creates new risks.
It should not become a convenient explanation for economic problems that existed before it.
STREAMING PLATFORMS ARE NOT ABOVE SCRUTINY
On 22 April 2026, the Texas Attorney General opened an investigation involving Spotify, Apple Music, Pandora, Amazon Music and YouTube Music over alleged undisclosed financial arrangements affecting promotion, playlists or recommendation rankings. Civil investigative demands were issued to the companies.
That needs to be stated precisely.
It is an investigation.
It is not a finding of guilt.
Calling the companies criminal organisations on that basis would be inaccurate.
But the investigation is highly relevant because algorithmic visibility has direct economic consequences. A platform influences what listeners discover; discovery drives streams; streams influence revenue.
Control over visibility is therefore also control over economic opportunity.
ROYALTY DISPUTES REMAIN PART OF THE STORY
Spotify has also been engaged in litigation with the US Mechanical Licensing Collective concerning royalty calculations associated with subscriptions incorporating audiobooks.
Earlier claims suffered setbacks, but in September 2025 a federal judge allowed the MLC to file an amended complaint advancing further allegations concerning Spotify’s Audiobooks Access product and royalty calculations. Those allegations remain allegations, not findings of wrongdoing.
The significance is not that Spotify should automatically be declared guilty.
The significance is that the streaming economy is an exceptionally complex financial architecture in which platforms, labels, publishers, rights organisations and creators frequently have competing interests.
The musician often stands at the end of that chain.
THE MACHINE-MADE FLOOD IS REAL
None of this should minimise the new problem created by generative AI at industrial scale.
In July 2026, Deezer reported receiving approximately 90,000 fully AI-generated tracks every day, with AI-generated material exceeding 50 per cent of new daily uploads at peak periods in June.
Yet fully AI-generated music represented only around 1 to 3 per cent of actual listening on Deezer, partly because the platform excludes detected AI tracks from algorithmic recommendations and editorial playlists.
More strikingly, Deezer reported that up to 85 per cent of streams on fully AI-generated tracks during 2025 were identified as fraudulent and demonetised.
That is not somebody experimenting with a synthesizer.
That is industrial-scale content production.
A human singer cannot record 90,000 songs today.
A composer cannot finish 90,000 complete pieces tonight.
A band cannot produce 90,000 albums over the weekend.
Software can operate at a completely different scale.
That affects discoverability, streaming fraud, metadata, recommendation systems, copyright and the finite amount of attention available to music.
This is one of the most important distinctions between generative AI and earlier recording technologies.
The problem is not simply that the machine can create.
It is that the machine can manufacture cultural material at a scale human labour cannot physically match.
MUSICIANS THEMSELVES ARE NOT SIMPLY ANTI-AI
APRA AMCOS’s AI and Music Report complicates the argument further.
More than half — 54 per cent — of surveyed Australian and New Zealand music creators agreed that AI could assist the human creative process, and 38 per cent were already using AI in their work.
At the same time, 82 per cent were concerned that AI could damage their ability to make a living, while the report estimated that 23 per cent of music-creator revenue could be at risk by 2028, with cumulative projected damage of approximately A$519 million.
That combination is critically important.
A creator can recognise the extraordinary usefulness of artificial intelligence while simultaneously fearing the business model developing around it.
There is no contradiction there.
THE TOOL CAN BE USEFUL WHILE THE ECONOMIC SYSTEM SURROUNDING IT REMAINS DEEPLY PROBLEMATIC.
THE REAL QUESTION IS WHO CAPTURES THE VALUE
Follow the economic structure from beginning to end.
ARIA determines chart eligibility. Labels control catalogues. Publishers negotiate rights. Streaming services sell subscriptions. Technology corporations sell artificial-intelligence systems. Record companies negotiate AI licences. Government promotes AI investment. Government encourages adoption. Public agencies use AI. Consumers receive unprecedented access to culture.
And the person who wrote, performed or recorded the song may still struggle to convert attention into economically meaningful remuneration.
That is the contradiction.
The fundamental problem is not simply:
Can artificial intelligence make music?
The larger question is:
WHO CAPTURES THE ECONOMIC VALUE CREATED AROUND MUSIC?
CONCLUSION — SO, HAS AUSTRALIA ACTUALLY BANNED AI MUSIC?
After all of this, the argument returns to the headline that started it:
AUSTRALIA BANS FULLY AI-MADE SONGS FROM OFFICIAL MUSIC CHARTS.
Catching?
Certainly.
Complete?
Not remotely.
Australia has not banned artificial-intelligence music.
ARIA has established an eligibility boundary for its own charts. AI-assisted music remains eligible under specified circumstances.
At precisely the same time, the Australian Government is actively attempting to grow an AI-enabled economy, encourage commercial adoption, attract multinational investment and place generative systems inside government workplaces.
Government institutions openly contemplate or use artificial intelligence for writing, imagery, voice, design and marketing-related work.
Major music corporations are not rejecting AI either. They are negotiating mechanisms through which it can be licensed and monetised.
Spotify simultaneously protects its platform against AI spam and impersonation while constructing commercial AI products with major rights holders.
ARIA simultaneously insists upon human creativity at the centre of chart-eligible recordings while operating inside a recording economy whose revenue and distribution systems are overwhelmingly technological and digital.
Government simultaneously talks about protecting human creators while promoting AI precisely because it promises productivity and reduced manual effort.
None of this proves malicious intent.
But the contradictions are impossible to ignore.
The real argument is not:
HUMANITY VERSUS THE MACHINE.
It is about control, authorship, licensing, copyright, employment, visibility, remuneration and institutional power.
And that takes the argument directly back to the question posed only days earlier in When the Machine Enters the Studio:
WHERE DOES THE TOOL END, AND WHERE DOES THE CREATOR BEGIN?
Perhaps the answer cannot be found merely by examining the software.
The more revealing questions concern who directed the technology, who made the meaningful creative decisions, who selected and rejected the outputs, whose existing creative work contributed to the system, who authorised that use, who was displaced by it, who owns the finished result, who receives the revenue and who possesses sufficient institutional power to declare the result legitimate.
Artificial intelligence is a tool.
Sometimes an extraordinary one.
Sometimes a liberating one.
Sometimes an unsettling one.
Sometimes a technology capable of expanding human creativity.
Sometimes a technology capable of economically eliminating the human whose labour it imitates.
The morality does not reside inside the computer.
It resides in what governments, corporations, industries and individual creators choose to do with the technology — and in how the economic benefits and costs are distributed.
ARIA is entitled to establish eligibility requirements for the ARIA Charts. That does not make ARIA the final authority on what constitutes music.
Government is entitled to pursue technological development. That does not absolve government from examining what happens when productivity reduces the demand for human labour.
Spotify can legitimately point to billions of dollars entering the music-rights economy. That does not invalidate the independent artist asking why an individual return remains economically insignificant.
Record companies can legitimately demand copyright licences from AI businesses. That does not mean commercially authorised automation automatically represents the interests of every working musician.
Artificial intelligence deserves scrutiny.
ARIA deserves scrutiny. Government policy deserves scrutiny. Major record companies deserve scrutiny. Streaming platforms deserve scrutiny. Technology corporations deserve scrutiny.
And above all, the economics surrounding the human creator deserve considerably more scrutiny than they have historically received.
Burroughs might have looked at this whole magnificent contradiction and recognised one of his favourite subjects immediately: control disguised as inevitability, commerce disguised as principle, and institutions attempting to define the limits of a technology they are simultaneously learning to exploit.
Perhaps the machine entering the studio was never the most disturbing part.
The greater danger is discovering that nearly everybody else in the building has already worked out how to extract economic value from the machine — while the independent artist is still asking where an equitable share went.
RELATED RESEARCH — INFECTIOUS UNEASE RADIO – SUBTERRANEAN ZONE RADIO – VOICE IN THE VOID
Gordon Lawrence Taylor — When the Machine Enters the Studio: AI, Computer Music, Human Labour, Lost Recordings and the Price of Making a Song — Where Does the Tool End, Where Does the Creator Begin, and What Happens When the Machine Becomes Cheaper Than the Musician? Infectious Unease Radio – Subterranean Zone Radio – Voice In The Void, 20 August 2026.
Read WHEN THE MACHINE ENTERS THE STUDIO
Gordon Lawrence Taylor — How Would William S. Burroughs See AI? — Cut the Machine: The Cut-Up Method, the Third Mind and the Machine Age of Writing, Art, Film and Music. Infectious Unease Radio – Subterranean Zone Radio – Voice In The Void, 22 August 2026.
Read HOW WOULD WILLIAM S. BURROUGHS SEE AI? — CUT THE MACHINE
BOOKS, ACADEMIC RESEARCH, GOVERNMENT REPORTS AND MUSIC-INDUSTRY SOURCES
-
David Cope — Experiments in Musical Intelligence. A-R Editions, 1996. ISBN 978-0-89579-314-0. An important early book on computer modelling of musical composition and style.
-
David Cope — Virtual Music: Computer Synthesis of Musical Style. MIT Press, 2001; paperback 2004. Hardcover ISBN 978-0-262-03283-4; paperback ISBN 978-0-262-53261-7. A substantial examination of computational style, aesthetics and artificial musical creativity.
-
David Cope — Computer Models of Musical Creativity. MIT Press, 2005; paperback 2017. Hardcover ISBN 978-0-262-03338-1; paperback ISBN 978-0-262-53410-9. Directly asks whether computational systems can model — and potentially produce — creativity.
-
Christina Anagnostopoulou, Miguel Ferrand and Alan Smaill, eds. — Music and Artificial Intelligence. Springer, 2002. Print ISBN 978-3-540-44145-8; eBook ISBN 978-3-540-45722-0. Proceedings of the Second International Conference on Music and Artificial Intelligence.
-
Roger T. Dean, ed. — The Oxford Handbook of Computer Music. Oxford University Press, 2011. Print ISBN 978-0-19-979203-0; online ISBN 978-0-19-994023-3. Covers computer composition, synthesis, performance, cognition and the longer history of machine-assisted music.
-
Nick Collins and Julio d’Escriván, eds. — The Cambridge Companion to Electronic Music, 2nd ed. Cambridge University Press, 2017. Hardback ISBN 978-1-107-13355-6; paperback ISBN 978-1-107-59002-1; digital ISBN 978-1-316-45987-4. Especially useful for demonstrating that musicians have repeatedly absorbed disruptive technologies into creative practice.
-
Jean-Pierre Briot, Gaëtan Hadjeres and François-David Pachet — Deep Learning Techniques for Music Generation. Springer, 2020. Hardcover ISBN 978-3-319-70162-2; eBook ISBN 978-3-319-70163-9. Detailed technical analysis of deep-learning systems used to generate musical content.
-
Eduardo Reck Miranda, ed. — Handbook of Artificial Intelligence for Music. Springer, 2021. Hardcover ISBN 978-3-030-72115-2; softcover ISBN 978-3-030-72118-3; eBook ISBN 978-3-030-72116-9. A 994-page reference covering machine musicianship, generative music, expressive performance and AI-assisted creativity.
-
Maria Eriksson, Rasmus Fleischer, Anna Johansson, Pelle Snickars and Patrick Vonderau — Spotify Teardown: Inside the Black Box of Streaming Music. MIT Press, 2019. ISBN 978-0-262-03890-4. Critical investigation of Spotify’s algorithms, infrastructure and platform economics.
-
Eric Drott — Streaming Music, Streaming Capital. Duke University Press, 2024. Paperback ISBN 978-1-4780-2574-0; hardcover ISBN 978-1-4780-2099-8; eBook ISBN 978-1-4780-2787-4. Examines streaming economics, surveillance, click fraud, artificial attention and cheap music.
-
David Hesmondhalgh — Music Streaming around the World. University of California Press, 2025. ISBN 978-0-520-42259-9. International research on Spotify, Apple Music, YouTube and streaming’s uneven cultural and economic effects.
-
Jeremy Wade Morris — Selling Digital Music, Formatting Culture. University of California Press, 2015. Hardcover ISBN 978-0-520-28793-8. Traces music’s transformation from physical recording to files, metadata, platforms and algorithmically managed digital commodities.
-
Australian Government — National AI Plan. Department of Industry, Science and Resources, 2 December 2025. Australia’s central whole-of-government AI strategy, built around investment, adoption, skills, public services and risk management.
-
Jobs and Skills Australia — Our Gen AI Transition. 2025. Major Australian labour-market research covering productivity, changed occupations, entry-level work and early evidence of displacement.
-
United States Copyright Office — Copyright and Artificial Intelligence, Part 2: Copyrightability. January 2025. Distinguishes AI-assisted human authorship from output determined primarily by generative systems.
-
UK Intellectual Property Office — Music Creators’ Earnings in the Digital Era. Government-commissioned research on musician income and the economics of streaming.
-
ARIA — “ARIA Charts Set Eligibility Rules for Recordings Made With AI” and “AI and the ARIA Charts — What You Need to Know.” Primary documents governing the Australian chart change effective 31 August 2026.
-
IFPI — Global Principles for the Eligibility of Recordings Developed Using AI in Official Music Charts. 30 July 2026. International recording-industry framework underlying ARIA’s Australian changes.
-
APRA AMCOS — AI and Music Report. 19 August 2024. Reports both significant creator adoption of AI and serious projected threats to music-creator income.
-
Spotify — Royalties Guide / Track Monetisation Eligibility. Primary source confirming the 1,000-stream annual threshold for inclusion in Spotify’s recorded-music royalty pool.
-
Spotify — Loud & Clear 2026 / 2025 Music Industry Payouts. Reports more than US$11 billion in 2025 music-industry payments and the distribution of artists across royalty-generation thresholds.
-
Music Business Worldwide — Spotify Loud & Clear 2026 analysis. Particularly useful because it distinguishes royalties generated from what musicians ultimately take home after labels, distributors and publishers.
-
Deezer — AI Music Upload Report, July 2026. Primary source for the reported 90,000 fully AI-generated tracks arriving daily and the extraordinarily high proportion of fraudulent streams associated with synthetic uploads.
-
ABC News — Josh Fawaz / Like a Prayer AI investigation and ARIA rule coverage. July–August 2026. Contemporary reporting on the recording that helped bring the Australian chart issue into public view.
-
Texas Office of the Attorney General — Investigation into Major Music Streaming Platforms. April 2026. Investigation concerning alleged undisclosed promotional arrangements involving Spotify, Apple Music, Pandora, Amazon Music and YouTube Music; an investigation, not a finding of wrongdoing.
-
Mechanical Licensing Collective v. Spotify USA Inc. US federal litigation concerning Spotify’s audiobook-related royalty calculations. Allegations remain subject to litigation and should not be represented as established misconduct.