For the last few years, the defence of AI in music has been pretty predictable. ‘Artists’ have cried to the detractors of the encroaching digital dystopia that it’s just another tool: it democratises access, lets people without money create, and helps independent artists compete. It gives the kid without a studio a shot at making something radio-ready enough to survive the algorithmic slaughterhouse.
And yeah, it does do all those things. So it is easy to see AI as the next step forward from synths and Auto-Tune instead of a complete moral collapse. The problem is that generative AI has moved far beyond the small-tool defence. It is no longer sitting politely in the corner helping someone clean up a vocal stem. It is in the song, the press release, the album cover, the gig poster, the biography, the fake artist profile, the fake streams, the fake connection. The whole ecosystem is being padded out with slop while artists are told to be grateful for the efficiency.
Now the ecological bill is starting to land. MIT Technology Review’s energy analysis found that generating one five-second AI video using a leading open-source video model used around 3.4 million joules, equivalent to running a microwave for over an hour. That estimate does not mean every AI use carries the same cost, and text prompts are far less energy-intensive, but it makes the scale of multimodal AI feel suddenly obscene when you picture a musician generating disposable micro-content for a single Instagram push.
So can anyone still defend this as harmless?
We’re already beyond a synthetic saturation point
The music industry has already passed the stage where AI can be treated as a strange novelty. Deezer reported in July 2026 that AI-generated tracks had exceeded half of all daily new music uploads to the platform, reaching roughly 90,000 AI-generated tracks per day at a June peak. Earlier in the year, Deezer said AI tracks were already around 44% of daily uploads, up from 39% in January.
Spotify has also been forced into defensive mode. In September 2025, it said it had removed more than 75 million “spammy tracks” over the previous 12 months during the explosion of generative AI tools, while also supporting an industry standard for disclosing AI use in music.
SubmitHub’s response has been more dramatic because it hits its own pocket. The platform’s new AI policy says it is blocking pure AI submissions and expects that decision to cut off around 10% of its revenue stream, a move it openly admits runs against normal business logic. The line in the sand is blunt: it should not be as easy as pasting lyrics into a machine and collecting a finished song from the other end.
SubmitHub exists inside the same independent ecosystem where AI defenders often claim to be fighting for accessibility. Here is a company built on submissions, curators and exposure deciding that the short-term money is not worth poisoning the well. That is not technophobic hysteria. That is market-level alarm from people who can see the pipes clogging.
The “just another tool” argument has started to rot
Some prominent musicians and producers have defended AI as a creative aid. Jean-Michel Jarre has urged the music and film industries to embrace AI, arguing that it can be used to create future forms rather than fake imitations. Rick Rubin has described AI as another tool for artists trying to realise a vision, while Diplo has spoken far more brutally about musicians needing to adapt.
There is a difference between using AI to separate a noisy stem, generate a rough visual sketch or experiment with sound design, and outsourcing the soul of the work to a predictive machine trained on the labour of others. That distinction is where the conversation should be, but AI culture rarely stays there. It slides from assistance into substitution, then tries to wear the language of empowerment while hollowing out the very disciplines it claims to democratise.
The “great equaliser” line also deserves suspicion. Equalising access to creation sounds noble until you realise the market is being flooded at a scale no human can compete with. If thousands of AI tracks can be generated, uploaded, playlisted, tested, renamed and re-uploaded with almost no friction, the independent artist has not been liberated. They have been buried beneath synthetic abundance.
It is a con to tell musicians that talent has finally been democratised when the outcome is a marketplace where attention becomes even more scarce, originality becomes harder to verify and human labour is expected to compete with infinite output made by people who have written a prompt and called it vocation.
The planet cannot keep underwriting throwaway content
The environmental concern is no longer a niche complaint from people who hate computers. The energy use of AI, especially video and image generation at scale, has become part of a wider anxiety around data centres, electricity demand and water use. Even OpenAI’s Sam Altman has publicly described a future where intelligence becomes a utility “like electricity or water”, bought on a meter, while also talking about making intelligence “too cheap to meter”.
That language should make every creative person sit up a bit straighter. Intelligence as a metered utility is not just a clever bit of infrastructure rhetoric. It recasts thought, writing, invention and problem-solving as something bought from a supplier. If that future arrives clean, fair and abundant, its defenders will call it progress. If it arrives through energy-hungry systems, copyright exploitation and mass professional displacement, then the “democratisation” story starts to look like PR sprayed over a land grab, bringing us back to the climate crisis.
The World Food Programme warned in 2026 that strengthening El Niño conditions could become one of the strongest on record, with impacts on food security in vulnerable regions likely to be felt into 2027 through future harvest cycles. The FAO and WMO have also warned that extreme heat is already disrupting crop yields, livestock health, water availability and rural livelihoods.
No, one artist generating one AI cover will not single-handedly collapse the wheat harvest. That kind of argument is too crude to be useful. But culture is made of normalised behaviours. Once AI visuals, AI bios, AI videos and AI tracks become standard promotional practice, the cumulative demand becomes harder to wave away. The same musician posting climate concern one week and generating disposable AI assets the next is not a monster; they are a symptom of a society that has been trained to value convenience until the invoice arrives.
Two realities are now fighting for the future of music
There are two realities playing out at once. In one, AI is thrilling, frictionless, democratising and inevitable. A teenager with no budget can make artwork, videos, songs, scripts and campaigns in a bedroom. An artist with no team can look like a label priority. The old gatekeepers lose some power, and that feels, understandably, intoxicating.
In the other reality, AI is an accelerant for fakery, redundancy and ecological strain. It rewards volume over experience. It makes the prompt writer feel like a creator while the actual training data carries the ghosts of artists who never consented. It gives platforms more cheap material, labels more leverage and audiences more reason to distrust what they are hearing.
The split is why the debate feels so vicious. AI is not merely a music-production issue. It has become a proxy for class, access, labour, climate, copyright, taste and dread. The aspiring AI artist sees possibility. The working musician sees erasure. The tech evangelist sees abundance. The songwriter sees their life’s work fed into a machine that may one day undercut them with a cheaper imitation.
Even the anti-AI side has to stay honest. Not every use of machine learning is the same. Restoring old audio, isolating stems, improving accessibility tools and helping disabled artists navigate production barriers should not be flattened into the same moral category as mass-generating soulless tracks to farm streaming royalties. The problem is not nuance; the problem is that the industry tends to use nuance as a loophole until the loophole becomes the business model.
Conclusion
Music made by humans has never been pure. It is commercial, compromised, derivative, overproduced and frequently annoying. But at its best, it still carries the voltage of someone trying to say something from inside a body, a room, a life, a debt, a desire, a humiliation, a breakthrough. That charge cannot be reduced to convenience without losing the whole bloody point.
If intelligence becomes something we buy on a meter, maybe creativity will become the last place where refusing convenience feels radical. Not because technology is evil, but because a culture that outsources too much of its feeling eventually forgets how to tell the difference between expression and output.
–
Article by Amelia Vandergast
Artwork by Zdzislaw Beksinski
No Comments