Universal urges Apple and Spotify to stop training AI models to protect music copyright

12 Apr. (Portaltic/EP) –

Universal Music Group (UMG) It has urged streaming music platforms such as Apple Music or Spotify to block the training of artificial intelligence (AI) models, because they would be violating the copyright of the songs they use.

The record label has indicated that it believes that certain AI systems “could have been trained with content protected by copyright” without obtaining the necessary consents and without paying financial compensation to the creators of said works.

The record label assures that it will not hesitate to “take measures” to protect the rights of its artists, according to some emails to which he has recently had access Financial Times.

In this way, it has asked the companies dedicated to these streaming content services to block generative AI developers from accessing their music catalogs.

“A lot of the generative AI is trained on popular music. You could say: I want to write a song with Taylor Swift-like lyrics, Bruno Mars-like vocals, and that resembles a Harry Styles song. The result you get is because the AI ​​has been trained on the intellectual property of those artists,” a related source commented.

From this newspaper they remember that Google already has a service that generates music from text, MusicLM, trained with a data set of 280,000 hours of music and that, for the moment, the company has not launched for a “risk of possible misappropriation of creative content”.

This is because researchers from the technology company would have discovered that around one percent of the music that this AI generated was an exact replica of the work with which it had been trained, that is, of copyrighted material.

It should be remembered that this is not the first time that professionals in the artistic sector have shown their concern about the possibilities of artificial intelligence and its ability to circumvent certain limitations. willing to protect copyright of its creators.

For example, last January a group of artists sued three companies dedicated to digital art –Stability AI, DeviantArt and midjourney– for infringing copyright in the development of artistic works created by AI with the Stable Difussion tool.

As a consequence of this problem, researchers from the University of Chicago created a solution called Glaze shortly after, which seeks to protect the works made by artists and prevent them from being end up training artificial intelligence models.