A digital rights group has claimed to have harvested a vast number of tracks from Spotify and is getting ready to release them online.
Possible Impetus for Artificial Intelligence Training
Observers said the seeming leak could aid AI companies searching for data to develop their technology.
An entity called Anna’s Archive reported it had scraped 86 million audio tracks from the platform and 256 million rows of metadata such as artist and album names. The company, which carries more than 100 million songs, confirmed that the breach did not constitute its entire inventory.
“We have identified and deactivated the malicious user accounts that were involved in unlawful scraping,” stated a Spotify representative.
The Stockholm-based service, which has more than 700 million users worldwide, added that an investigation into unauthorised access identified a external actor scraped public metadata and employed illicit methods to bypass DRM protections.
Archive's Declared Goal and Concerns
Spotify does not think the audio acquired by Anna’s Archive has been made public yet. Anna’s Archive, which is known for offering links to pirated books, wrote in a blog it wanted to create a ‘historical repository’ for music.
The group asserted the music files accounted for 99.6% of all music streamed to by its listeners and would be shared via ‘peer-to-peer networks’, a means of exchanging large data files online.
“Naturally the service doesn’t have all the music in the world, but it’s a fantastic start,” commented Anna’s Archive, which frames its purpose as “safeguarding humanity’s knowledge and heritage”.
AI Training and Intellectual Property Battleground
A composer and campaigner for defending artists’ rights, remarked the leaked music would likely be used for training AI models.
“Training on illegally sourced material is sadly widespread in the tech sector, so this stolen music is highly likely to be used instructing AI models,” he said.
The group's website makes references to LibGen, a vast online repository of copyrighted books that has reportedly been used by Meta to train its AI models.
The co-founder of an AI startup posted on LinkedIn that members of the public could in theory “create their own private free version of the service”. The executive said it could also enable tech companies to “educate models on contemporary music at scale”.
Copyright has become a key issue between creatives, writers and creatives on one side and AI companies on the other.
Policy Considerations and Reactions
Across Britain, creative professionals have protested against a government proposal to allow AI companies use copyright-protected work without consent unless the owner opts out. Almost every respondent to a government call for evidence on the idea has backed artists’ apprehensions.
A government minister for science, research and development informed parliament this month there was “no clear consensus” on the matter, noting that ministers would “take the time to get this correct”. The administration has pledged to make regulatory plans on AI and copyright by March 2025.
