Recently, I found myself facing a decision I had not anticipated.
I had been approached to sing an Odia bhajan — ‘Aahe Nila Shaila’ — for a band called Trilok. I have always believed that artistes who have acquired visibility, like me, should occasionally lend that visibility to younger musicians. An Odia bhajan, new talent, a cultural collaboration in the month of the Rath Yatra — it seemed like something I would naturally want to support.
Then I did a little research.
I discovered that Trilok was not a band in the sense in which most of us understand the word. Its members were virtual identities. It now has some men with painted faces to play live stages. The project had been developed as an AI-led “spiritual rock band” by a talent-management and media company, not by four musicians who had met, rehearsed, disagreed, failed and finally found a sound together.
I would have appreciated being told this at the very beginning. The fact that this was an AI-generated band with fictional identities was not a minor production detail. It was central to the nature of the invitation and materially changed my decision about whether I wished to associate my voice with it. I gave my answer the same night. It did not come easily, because I am not opposed to technology. In fact, I had to examine my own recent choices before arriving at it.
Every generation of musicians has feared some new machine. When synthesisers entered studios, many instrumentalists believed their livelihoods would disappear. The anxiety was understandable: Why hire an orchestra if a keyboard could imitate one? Why call a drummer when a machine could keep perfect time?
But the synthesiser did not bring music to an end. It became another instrument. Drum machines did not eliminate rhythm; they altered its vocabulary. Sampling, digital recording, Auto-Tune and home-production software all provoked similar alarm. Each was eventually absorbed into music.
So it would be simplistic to declare that any use of AI is automatically an attack on art.
I recently released ‘Ghane Bhadra’, an original song whose video combines a live green-screen shoot of me performing with AI-assisted imagery. I am bodily present in the video. I sang the song. Human musicians played on it. AI helped us enlarge the visual canvas. It did not invent the artiste, replace the performance or conceal the people who made the music. That distinction matters to me.
Independent musicians in India work within an increasingly punishing economy. We write and produce original songs, pay for recordings, musicians, mixing, mastering, styling, shoots, editing, publicity and distribution without the machinery of a major film studio or record label. We then place that work on social-media platforms that depend on our content to hold audiences but do not pay most of us meaningfully for it. A new technology that allows us to tell a more ambitious visual story without waiting for a label’s permission or budget can feel liberating.
This is why I had to ask myself an uncomfortable question: Was I being inconsistent? How could I use AI in my own video and then refuse to sing for an AI band?
The answer became clearer when I stopped treating all uses of AI as though they were the same.
There is a meaningful difference between an artiste using a machine and a corporation using one to manufacture the appearance of artistes. One begins with human authorship and uses technology to extend its reach. The other can create an entire synthetic identity around which a catalogue and business may be built. One helps a financially constrained artiste complete a labour of love. The other can produce material at a velocity no human musician can possibly match. One acknowledges the people at the centre. The other risks making them invisible.
When I examined the Trilok project, what troubled me was not simply that the members were digital avatars. It was the larger proposition. Devotional music is among the most enduring and commercially powerful musical traditions in India. The songs come with faith, familiarity and an audience already attached to them. A synthetic act built around beloved bhajans can rapidly create a catalogue without undergoing the slow, uncertain process through which musicians develop a voice.
Except for the guest collaborator, a known name used presumably to expand reach and gain credibility, all Trilok uploads just credit “Trilok” as the performer. That makes the question of authorship even more important. The point is not that no human work exists behind an AI output. Of course it does. But if human beings perform the labour while fictional identities receive the fame, where will value eventually accumulate?
And what happens when a recognised human singer is brought in?
My voice could give such a project something an avatar cannot manufacture: History, credibility, cultural location and emotional trust. My participation could tell listeners that this synthetic band had been accepted by a working artiste. Even unintentionally, I might help normalise a model that could ultimately reduce the space available to the world I come from.
That was the ethical and moral line. Would I cross it?
There is another unresolved question. Generative music systems learn from large bodies of existing material. Around the world, creators are asking whether copyrighted work has been used for training with consent, transparency or compensation. The legal position is still contested. Technology has always learned from human knowledge. Artistes, too, learn from other artistes. But artistes do not ingest a million recordings at machine scale and then produce a competing catalogue within days. Scale changes the ethical character of the act.
I do not believe the answer is panic, prohibition or romantic speeches about returning to some pure pre-technological past. But artistes need to ask harder questions before lending their names and voices to synthetic acts. Who owns the resulting work? Who is credited? Was the training material licensed? Can a voice be cloned later? Is the human guest artiste helping build their own replacement? Is technology reducing the cost of expression or reducing the need to compensate creators? Is the audience being told clearly what it is hearing?
And the final question: Where is the human being in this, and are we being invited to see them?
After thinking through, I knew I could not participate.
My use of AI in ‘Ghane Bhadra’ may still be questioned. That is fair. Artistes should not demand exemption from scrutiny simply because our intentions are good. I am also learning where my boundaries lie. But this experience clarified one of them. I will use technology to help human expressions travel further. I cannot knowingly lend my voice to make the human artiste disappear.
That is not fear of the future. It is a choice about the kind of future I am willing to help create.
The writer is a singer, producer and live performer
