Generative AI (GenAI) is advancing rapidly and transforming creative industries. With its ability to generate songs in a few minutes, it raises questions about AI’s role in human creativity. How can the EU find a balance between supporting the development of technologies that drive innovation and protecting citizens from their social, ethical and legal risks?
In the run-up to the STOA workshop “Empowered by Music in the Age of Innovation” organised next week on 4 November at the European Parliament, the ESMH explored these issues with Dr François Pachet from Imagine All The People, and former director of Spotify Research Lab and Sony Computer Science Lab.
How might the increasing use of AI-generated tools affect musicians’ creative process? Will text generators, for instance, become integral to writing lyrics? How many of the top 100 songs in 2026 do you believe might contain AI-generated elements?
François Pachet: Generative tools infiltrate everything, often invisibly: rhyme and flow search, topline sketches, arrangement assistants, temporary vocal stems, mastering, source separation. Lyricists won’t use a “hit generator” but a version companion producing fifty variants, mixing registers, surprising a metric. By 2026, I’d bet that 60 to 80 percent of Top 100 songs will contain some AI component (from text suggestion to vocal denoising), most unacknowledged. The public will judge the audible effect, not the procedural purity.
The democratisation of music production tools has led to a surge in music creation, but has it also enhanced the overall quality of the music?
François Pachet: We have drastically lowered the cost of entry, but we have not magically increased talent. The quality distribution has flattened: many more mediocre tracks, and a few more excellent ones. Good tools amplify good signals when guided by a discerning ear; otherwise they homogenise. Value is shifting toward editorial taste (selection, curation, listening contexts) and the craft of singularity (sounds, processes, narratives), not toward the “magic master” button.
Will the EU regulations on data protection and quality control of AI-generated content ultimately benefit the music industry and other creative industries by setting new standards and making the EU a leader in responsible AI, or will it rather hinder innovation and competitiveness in these fields?
François Pachet: When lawmakers legislate without grasping the technical mechanics, they build bureaucratic machines that punish European innovators rather than the well-resourced foreign giants who can tick every compliance box. The AI Act embodies this: a defensive reflex, ideological polarisation, and technocratic compromises disconnected from real practice. The likely outcome is procedural inflation that chills experimentation in music and cultural industries, without any clear benefit for artists.
“Quality” cannot be decreed; it emerges from living ecosystems where people can try, fail, and iterate. If Europe wants to lead in “responsible creativity,” it should invest in open, pragmatic standards, light-weight traceability, auditable models, and clear usage rights, not in a regulatory Maginot Line.
In the long run, will AI primarily serve as a tool to enhance artistic expression, or will it drastically change the role of the ‘human artist’?
François Pachet: AI doesn’t “feed on” artworks to regurgitate them; it learns statistical similarities. It’s an instrument that enlarges the possible gestures. The human role shifts toward direction of meaning: framing an aesthetic, imposing constraints, recognising the precious accident. Like electricity, the sampler, or the DAW, these technologies redefine authorship but do not abolish it; they strengthen those who can choose, sign, and own their decisions.
What emerging technology or innovation could surpass and replace GenAI as the next tool for human creativity, making GenAI outdated or obsolete?
- Structure-aware (neuro-symbolic) models:generators that handle explicit musical or poetic constraints (form, voice, harmony) rather than mere local correlations. A move from style-transfer to craft-transfer.
- Real-time co-creative agents: systems that play with you: listen, react, remember, and learn your taste across sessions instead of isolated prompts.
- Integrated traceability and licensing tools:robust watermarking, edit logs, programmable rights. Less glamorous, but the true catalyst of sustainable creative ecosystems where contributions are recognised and value can be shared.
François Pachet: These directions could make today’s correlation-based GenAI obsolete, replacing it with composition-oriented machines built on structure, interaction, and responsibility.
Useful link:
• STOA workshop ‘Empowered by Music in the Age of Innovation’
• Webstream link