(Paris) Universal Music Group obtained this week the withdrawal of a rap song generated by artificial intelligence and imitating the voices of artists Drake and The Weeknd, but this one did not necessarily infringe copyright, according to Andres Guadamuz, professor of British law.
“You can’t copyright someone’s voice,” Guadamuz, who teaches at the University of Sussex in Britain, told AFP.
Created by a user named @ghostwriter, the track Heart On My Sleeve went viral on the social network Tiktok and had been listened to millions of times on Spotify, Apple Music and other platforms, before being removed after a request from the major.
According to the professor, like a melody, lyrics or other elements of a song, the timbre of a voice could be protected “but that would be problematic”, because “copyright protects the expression of a idea, what the voice really is not”.
However, this specialist believes that Universal probably obtained the withdrawal of the song by relying on copyright, which benefits from proven procedures with streaming platforms.
“Most of the time these issues aren’t resolved by law, but just because the record companies are complaining to the platforms and it’s easier for them to comply,” says- he. Especially since Universal Music Group or Sony are, for example, minority shareholders in Spotify.
Other laws, however, may already protect a musician against artificial intelligence copies and, if an artist has a particular personality or voice, these characteristics may be protected by publicity rights. in the USA).
In 1988, actress Bette Midler won a lawsuit against Ford, who had used an imitation of her for an advertisement. Singer Tom Waits won a similar case against potato chip giant Frito-Lay in 1993.
But the application of this right is “very random”, notes Andres Guadamuz, with some countries interpreting it more strictly than others.
The legal battle could finally be over how artificial intelligence programs are trained. In the case of Drake and The Weeknd, they necessarily needed many works of these two artists to be able to imitate them.
“You need to copy the music to train AIs and this unauthorized copying could infringe copyright,” says the professor.
“Defenders will say, however, that this is fair use. They use these copies to train the machine, teach it the music, and then they delete them. We will have to wait to see how these cases will be judged,” he adds.
In the meantime, it will be difficult to stop the avalanche of content generated by artificial intelligence. Just this week, an album impersonating Oasis frontman Liam Gallagher has been spreading across social media.
“Groups are going to have to decide if they want to sue this because copyright cases are expensive,” Guadamuz said.
“Some artists might also lean on the technology and start using it themselves, especially if they start to lose their voice,” he anticipates.