28% of Music Delivered Daily to Deezer is AI-Generated: Creation Sounds the Alarm

Faced with the explosion of AI-generated music content, rights holders are taking up arms. Deezer warns of a massive phenomenon, pointing fingers at tools like Suno and Udio, and is implementing advanced detection systems. However, the music industry, like the entire cultural sector, is entering a tense zone.

It's a figure that sent shockwaves through the music world: according to Deezer, 28% of the tracks delivered daily to its platform are "entirely generated by artificial intelligence". That's more than 30,000 tracks per day, often without explicit mention, clear credits, or declared human intervention.

However, these AI tracks represent only a small portion of actual listens: about 0.5% of total streams, according to the platform. This figure remains concerning for the music ecosystem: it reveals a covert industrialization of music production, largely operated by automated systems.

The Tools in the Crosshairs: Udio and Suno

Two names keep coming up: Udio and Suno. These generators allow for the creation of a complete song in seconds, from a simple phrase or musical style. Sung vocals, arrangements, rhythm, mastering: everything is synthetic.

🔍 Universal Music recently filed a lawsuit against Suno, accusing it of using thousands of recordings from its catalog to train its AI, without authorization or compensation. The complaint filed in the United States mentions massive copyright infringement and demands transparency about the datasets used.

For Udio, the same logic applies: the tool is remarkably effective, but its opacity regarding training sources is causing growing concern among rights holders.

In response to this surge, Deezer has developed in-house technology to identify AI-generated tracks, particularly those produced by Suno and Udio. The platform claims to detect these contents with a reliability rate close to 99.8%.

The system relies on audio signal analysis: AI-generated tracks leave statistical artifacts imperceptible to the human ear but identifiable by machine learning models. Deezer also uses metadata and delivery patterns to refine its detection.

The platform claims a generalizable approach, capable of recognizing tracks generated by yet unknown tools, without needing a specific training set.

Measures Against 100% AI-Generated Music

Once detected, these tracks are not removed but deactivated from several monetization circuits:

• They no longer appear in algorithmic recommendations.
• They are excluded from Deezer's editorial playlists.
• Suspicious or fraudulent streams (bots, automated clicks) are identified and excluded from royalty calculations.
• An automatic filter is implemented to prevent these AI tracks from disrupting listening or diverting undue revenue.

In short, Deezer seeks to preserve the platform's economic balance by preventing artificial intelligence from diverting artist compensation to anonymous producers or content factories.

A Period of Open War with Rights Holders

These technical measures come amid an increasingly tense climate. Creation is in open war against generative AI. Authors, performers, management companies, producers... all denounce a system where their works are used without authorization or compensation to generate monetized content elsewhere.

In France, organizations like ADAGP, SACEM, ADAMI, and FIM are stepping up. A manifesto signed by over 30,000 artists speaks of a "looting" of cultural works by AIs and demands:

• Transparency on the training datasets used by AI models;
• Implementation of a right to opt-out or, preferably, prior consent (opt-in);
• Compensation for artists whose works have been used;
• Mandatory marking of generated content (watermark, metadata, explicit mention);
• Human quotas in recommendation algorithms to preserve artistic diversity.

The strong symbol of this mobilization: the release of a silent album (Is This What We Want?) by hundreds of musicians, to make the silence of a world deprived of human creation heard.

Deezer's case may be the beginning of broader regulation. Streaming giants will have to choose their side: encourage automation at the risk of stripping art of its meaning... or defend a creative ecosystem based on humanity, intention, emotion.

On one side, tools capable of producing 100 songs a day. On the other, artists reminding us that music is not just a sequence of sounds—it's a voice, a story, an emotion.