This year’s Sónar+D felt like stepping into a line between two futures: one where AI becomes a wildly inventive musical collaborator, and another where the same infrastructures quietly fuel surveillance and warfare.
Over two dense days at the new Llotja de Mar venue in central Barcelona, Sónar’s innovation lab brought together artists, researchers and technologists to ask what kind of world we’re actually building with these tools.
What emerged was a striking tension: on one side, genuinely optimistic experiments in human–machine collaboration; on the other, a cold, forensic look at how the same systems can scale violence when they’re steered by the wrong incentives and the wrong people in power.
Evicshen: Chaos at the End of Day One
Thursday at Sónar+D closed not with a sleek AI demo, but with a bodily, chaotic, and defiantly analog performance by Evicshen (Victoria Shen), who took over Stage+D from 20:00 to 20:45.
Shen is a sound artist, experimental performer and instrument‑maker based in San Francisco, known for building her own electronics and using “chaotic sound” to short‑circuit our usual expectations of rhythm and melody.
Her practice leans heavily on analog modular synths, sculptural sound objects and self‑built devices including her “needle nails,” acrylic nails embedded with turntable styli that allow her to play multiple grooves of a record at once.
At Sónar+D, that approach became a kind of counter‑spell to the smooth promise of generative AI. In a festival where “intelligence” is often quantified in model parameters and inference speed, Evicshen’s set insisted that raw feedback, unstable circuits, and the performer’s own body are also computational systems.
The result felt like a reminder that, even in an AI‑heavy programme, some of the most radical sonic experiments still come from humans wrestling with stubborn hardware, electricity, and physical space.
Sonic Experiments with Daito Manabe and Google DeepMind
Japanese polymath Daito Manabe and Google DeepMind’s team leaned all the way into machine learning. Their “Sonic Experiments with Daito Manabe and Google DeepMind” session dug into the integration of DeepMind’s Lyria – a neural audio model – into Manabe’s new AV show, which debuted at Sónar 2026 after a collaboration that began at Sónar+D 2025.
On stage, Manabe joined Jeff Chang from Google DeepMind to unpack what it actually means to co‑write music with a model rather than simply generate endless loops in the background.
The conversation moved beyond technicalities like prompts and pipelines, and into the messier territory of authorship and trust: when does an AI feel like an instrument, and when does it start to feel like a genuine collaborator that pushes you somewhere you couldn’t have gone alone?
They spoke about moments when Lyria’s output felt uncannily intuitive – surfacing harmonic ideas that fit Manabe’s language, and moments when the system’s tendencies clashed with his aesthetic, forcing decisions about where to draw a line and reassert human intent. Human taste will prevail in the era of AI, what tools we use, doesn’t matter so much to Manabe.
The session captured an overall mood among many artists at the festival: not a fearful rejection of AI, but a pragmatic, almost playful attempt to put these tools to work as collaborators rather than replacements.
Dark Forests, Data Exhaust and the “Kill Cloud”
The optimism at Sónar+D was not naive though, and nowhere was that clearer than in the conversation between Joana Moll and Wassim Z. Alsindi who drilled into the uncomfortable reality that the same infrastructures making large‑scale AI possible – surveillance capitalism, ad‑tech, data extraction – are also entangled with military and intelligence operations.
Moll’s work exposes the invisible infrastructures of the internet: from the ecological cost of a single click to the opaque data pipelines and business models behind the platforms we treat as neutral utilities.
In Berlin, she has already connected online advertising technologies to what whistleblowers Lisa Ling and Cian Westmoreland describe as the “Kill Cloud” – a distributed weapons system in which drones, sensors, data centers and decision‑support software form a continuous network for identifying, tracking and targeting people.
As Moll explains, the concept captures how an enormous, data‑hungry apparatus originally justified in terms of “security” and “precision” strikes can end up normalizing remote, asymmetric warfare, with the web itself and ad‑tech infrastructure acting as part of that machine.
In their Sónar+D dialogue, Moll and Alsindi turned that lens back on everyday internet use: what happens when the same ad‑tech systems that target us with personalised banners are also capable of tracking intelligence agents, or civilians on the ground in conflict zones, based purely on their data exhaust.
Alsindi, co-founder of 0xSalon, brought in perspectives from cryptography and digital culture, probing how power, identity and “truth” are shaped in an online environment where every interaction can be logged, monetised and potentially weaponised.
The takeaway was clear – we cannot talk about AI as a neutral creative tool without also acknowledging the corporate and geopolitical architectures that underpin it – from advertising platforms to defense contractors – and the very real possibility that these systems are already being repurposed for information warfare, autonomous weapons and automated decision‑making about life and death.
Two Futures in the Same Room
Threaded through all of this was a very Sónar‑specific energy: a community of artists, coders, theorists and festival‑goers who are genuinely excited about what AI can do on stage, yet increasingly unwilling to ignore what it can do in the hands of unaccountable power.
Sónar+D didn’t resolve that tension, at its best, it made that duality impossible to ignore – and quietly asked everyone in the room which future they’re actually helping to build.