The era of music without musicians: when “to say is to create.”

According to figures released by the streaming platform Deezer (Le Monde, August 3, 2025), nearly 20,000 new music tracks created by tools such as Suno are uploaded every day. The problem is that these new songs—whose growing volume tends to drown out the rest of musical production—are claimed by artists who do not exist. The newspaper Le Monde reportedly identified more than fifty “bands” likely generating their music through AI. Although platforms such as Deezer use tools to detect AI-generated tracks, it currently appears impossible to identify them all.

A paradoxical fact, symptomatic of the current generative AI revolution: the young American start-up Suno, founded in December 2023 and claiming more than 50 million monthly visits, is said to have been trained on millions of songs while largely disregarding the elementary rules of intellectual property law. Hence the lawsuits it currently faces for copyright infringement. In its paid version, Le Monde notes, Suno grants users full commercial rights over the songs they generate through the application, while claiming to embed a “watermark” (a pattern integrated, visibly or invisibly, into a digital file—such as an audio file—intended to indicate ownership and protect copyright) that can only be detected by machines.

According to Suno users, it has now become very easy to create personas—a term commonly used in digital marketing to designate the fictional representation of a user, usually constructed from behavioral, psychographic, and sociodemographic data—of singers with unique vocal signatures that can be reused from one track to another. As Deezer’s innovation director explains, “thanks to AI, it is now possible to generate catalogs that are completely realistic. (…) You just go on Suno and you can create an entire catalog that sounds like the Beatles” (Le Monde, August 3, 2025). All musical styles are represented: blues, country, French pop, fictional metal bands, electronic music, or even religious music.

Generative AI thus seems to confirm that it is no longer necessary to possess any technical talent—drawing, painting, or musical skill—to suddenly present oneself as an illustrator, painter, or singer. It is as if the creative process itself had definitively disconnected ideation from execution. A movement already initiated, to a certain extent, by the invention of photography in the first half of the nineteenth century: a mechanical means of instant reproduction of reality emerged, forcing painting to question its mission of faithfully representing the real and to free itself from the obligation of realism.

Users of Suno, moreover, though not musicians, claim the songs they create as genuine artistic creations precisely because they wrote the lyrics and ensured the creative direction (through prompts). In other words, music—and art in general, at least when it takes the form of pictorial or even cinematic creation—tends to become what linguists and philosophers of language call performative acts. That is to say, the very act of producing a work of art—such as a musical track created with Suno—would lie entirely in our capacity, as speakers, to utter statements (prompts). To borrow the title of a famous work by the philosopher of language John Austin: “to say is to do.”