AI and Music: Are We Just Repeating Ourselves?

Last week at SXSW London 2026, I sat in on a fireside chat that stuck with me. Paul Pacifico, CEO of the Saudi Music Commission, and Jeremy Silver, CEO of the British Library, took to the stage for a conversation billed around AI, ethics, and the future of music. What followed was a candid, wide-ranging conversation with two industry veterans who clearly care about music deeply, wrestling honestly with questions that don't yet have clean answers.

The line that landed hardest came early … AI, someone said, is like Ozempic. Everyone's on it. Nobody wants to admit it.

It got a laugh. But it also cut to something real.

So, here are comments, thoughts and elements I picked up on and wanted to share with you. All thought provoking and hopefully gets you thinking too! 

---

The Same Disruption, Different Decade?

Let’s face it, the music industry has been here before, we all remember home taping, napster, the value gap, etc. Each wave of technology arrived feeling like an existential threat, and each time the industry found its way through — battered, changed, but intact. The question hanging over the room was whether we're watching the same film again, just with better special effects.

It's a genuinely hard pattern to break. When the stakes feel existential, the instinct is to protect what you have. The internet years were bruising, and the eventual shift to streaming didn't come easily or cleanly. Now, with generative AI, the industry finds itself back at a threshold that feels uncomfortably familiar. 

The tools are more powerful, the creative implications deeper, and the desire to get it right this time (rather than spend a decade in reactive mode) feels palpable in conversations like this one, and in every AI discussion we've been part of across the industry over the past few years.

Are we in what one speaker called "a conspiracy of optimism"? Does the tech world genuinely care about ethics, or is ethics just a word that buys goodwill while the training data flows?

What AI Can and Can't Do (Yet)

One of the more grounded moments in the conversation was an honest assessment of generative AI's actual output. It's not uniformly impressive. It tends toward the polished and the generic — competent at pop, less so at the blues. There's something telling in that. The genres AI handles best are the ones most built on formula. The ones it struggles with carry more of what makes music human: feeling, friction, history, place.

Jeremy Silver made an observation that stayed with me. The British Library's instinct — perhaps uniquely — is to collect and not to judge. Would they archive a book written entirely by AI? In the UK, probably yes, because it's culturally interesting, regardless of the legal status of its authorship. 

The US position is starker, no copyright, no collection, and the divergence matters. We keep reaching for universal frameworks in AI governance, but we don't even have universal frameworks for collecting societies. Why would this be different?

The Questions That Actually Matter

The ethics conversation in AI and music often gets dominated by the loudest issues; copyright, voice cloning, training data. Those matter. But the panel surfaced some less-discussed tensions that I think deserve more airtime.

The talent pipeline problem.

When organisations cut junior roles in pursuit of AI-driven efficiency, they're not just saving costs today. They're removing the entry points through which the next generation of music professionals learn their craft. The people who would have spent three years doing the grinding, unglamorous work that AI now handles…they're not getting hired. In ten years, who will lead this industry? Your board wants efficiency. But who's thinking generationally?

The unpaid small claims problem.

Paul Pacifico raised something that doesn't get enough attention: the vast pool of unresolved royalty claims sitting in the system, particularly for small and midtier artists. The amounts per claim are often too small to be worth pursuing through legal or administrative channels. But for an emerging artist, that money (if it could be unlocked) could be the difference between continuing and giving up. This is actually a problem AI could help solve. Not the glamorous use case, but perhaps the most meaningful one.

The attribution problem.

There's still no reliable way to know when and where an artist's work has been used to train a model, or to compensate them accordingly. Until there is, the ethical framework everyone talks about building has a hole at its centre.

What AI Won't Replace

The panel ended somewhere that felt genuinely hopeful rather than reflexively.

Live music, one speaker argued, is the one thing we know is real. Post-pandemic, there's a renewed understanding that human presence, the shared physical experience of performance, carries something that can't be synthesised. AI isn't a competitor here. It's not even in the same category. 

And there's something broader in that. Communication isn't just information exchange. It's body language and breath and the specific energy of a room. Music at its most powerful operates in that same register. The question isn't whether AI can approximate some of that…it can, and increasingly well. The question is whether approximation is the point.

Where I Land

I left that room with more questions than answers, which is probably the right outcome for a genuinely complex topic. But a few things felt clear.

The music industry has both the most to lose and the most to offer in this conversation. More than any other creative sector, it has lived through disruption at scale and come out the other side with hardwon knowledge about what works, what doesn't, and what gets lost along the way.

The ethical frameworks will only be as good as the people building them! And right now, the people who care most about music need to be in those rooms, not just the people who care most about efficiency.

And if “AI really is like Ozempic”, something everyone's using, quietly, while pretending otherwise, then the more honest conversation starts with admitting that, and asking what we actually want from it.