Apple Music Told to Crack Down on AI-Generated Music: Report
- Global music giant Universal Music Group has asked Apple Music, Spotify, and other music streaming platforms to purge AI-generated music.
- UMG wants to prevent AI music generators from being trained on copyrighted music.
- Google created an AI model that could generate music from any text description but never released it.
Apple Music and other streaming platforms, including Spotify, have received requests from global music giant Universal Music Group to prevent AI technology developers from training music generation models on its copyrighted music — reports the Financial Times.
UMG controls almost a third of the global music market. The label has signed many big-ticket artists, including Taylor Swift, The Weeknd, and Elton John.
With the rise of AI music generators like ChatGPT creator OpenAI‘s Jukebox, UMG has grown concerned about AI models designed to artificially generate music that sounds like popular artists being trained on datasets that include songs the company and its artists own the rights to.
“We will not hesitate to take steps to protect our rights and those of our artists,” UMG wrote in emails to online music platforms in March, according to the Financial Times.
“We have become aware that certain AI systems might have been trained on copyrighted content without obtaining the required consents from, or paying compensation to, the rightsholders who own or produce the content,” UMG told streaming services like Apple Music last month.
UMG has already been trying to cleanse streaming platforms of “lower-quality” songs like ambient music and AI-generated tracks. The company has been sending takedown requests “left and right,” per one person familiar with the matter.
“This next generation of technology poses significant issues,” said a person close to the situation. “Much of [generative AI] is trained on popular music. You could say: compose a song that has the lyrics to be like Taylor Swift, but the vocals to be in the style of Bruno Mars, but I want the theme to be more Harry Styles. The output you get is due to the fact the AI has been trained on those artists’ intellectual property.”
Perhaps the best current example of this kind of technology is Google’s MusicLM, an AI model that can generate music from any text description. Google trained MusicLM on a dataset comprising 280,000 hours of music, but the tech giant never released it because of the “risk of potential misappropriation” of copyrighted content.
In a statement to the Financial Times, a spokesperson for UMG said:
We have a moral and commercial responsibility to our artists to work to prevent the unauthorised use of their music and to stop platforms from ingesting content that violates the rights of artists and other creators. We expect our platform partners will want to prevent their services from being used in ways that harm artists.