The news is not that AI art reached a music festival. It is that the audience is now asked to behave, not just watch.
Sonar+D 2026, the technology strand of Barcelona's Sonar festival, runs 18-19 June inside the neoclassical Llotja de Mar, foregrounding embodied robotics, AI interfaces and site-specific LED visuals, according to designboom. Volvox Labs' 'Astral Twin' pairs robotic arms that paint in dialogue with a giant LED wall, while Qs Ventures' 'A Voice from the Future' is a moss-covered phone booth inhabited by an AI persona. Several installations deploy AI tools, including Google DeepMind's Lyria 3, to generate soundscapes in real time.
The real signal is behaviour, not playback
It is easy to file this as AI art arriving at a music festival. The deeper move is a change in what the audience is asked to do. These are not screens to watch; they are systems to converse with, a robot that responds, a booth that answers, a soundtrack generated in the moment. The new default is behaviour, not playback, and that reframes where the work actually lives.
A looped video is authored once and shipped. A responsive installation is authored as rules: what it senses, how it reacts, what happens when two visitors interact at once, and what it does when no one is there. The deliverable becomes a behaving system with a fail state, not a finished file. Budget and risk shift from production toward logic, tuning and testing, the parts that never appear in the render, and the parts that decide whether the piece survives contact with a live crowd.
What a producer can do now
Ask for the interaction model before the visual. For any AI-driven or robotic piece, the questions that decide success are unglamorous: latency between input and response, behaviour under a crowd, how a generative element is bounded so it never produces something off-brand or unsafe, and who is on call when it drifts. Interactive art formats live or die on the feel of that response loop, and you can prototype the loop with a rough rig and one interaction long before committing to robotic arms or a bespoke booth.
Why it belongs in an experiential toolkit
Generative and responsive systems are reusable in a way fixed content is not. A behaviour tuned for one activation can be re-skinned for another brand, fed a different dataset, or bounded by new rules without rebuilding the rig. Treated as a capability rather than a one-off artwork, an AI-driven interaction becomes an asset you can redeploy, and a set of safeguards you only have to design once.
SensaLab is the white-label real-time 3D and immersive layer agencies and brands use to design and run interactive work under their own name, from the response loop itself to the guardrails that keep a generative system on-brand.