A digital rights group has claimed to have harvested millions of songs from the popular streaming service and is preparing to publish them publicly.
Potential Impetus for Artificial Intelligence Training
Industry experts said the apparent leak could aid AI companies looking for material to train their models.
An entity known as Anna’s Archive reported it had obtained 86 million music files from Spotify and 256 million entries of metadata such as performer and album names. Spotify, which carries more than 100 million songs, stated that the leak did not represent its entire inventory.
“We have located and disabled the nefarious user accounts that engaged in unlawful scraping,” stated a company spokesperson.
The Stockholm-based company, which has more than 700 million subscribers worldwide, further explained that an inquiry into unauthorised access identified a third party collected public metadata and employed questionable tactics to bypass DRM [digital rights management].
Group's Stated Goal and Worries
Spotify does not believe the audio acquired by Anna’s Archive has been released yet. Anna’s Archive, which is recognized for offering links to pirated books, wrote in a blog it wanted to create a ‘historical repository’ for music.
They asserted the music files accounted for 99.6% of all music listened to by Spotify users and would be distributed via ‘peer-to-peer networks’, a means of exchanging large data files online.
“Of course Spotify doesn’t have all the music in the world, but it’s a fantastic start,” said Anna’s Archive, which frames its purpose as “safeguarding humanity’s knowledge and heritage”.
AI Training and Copyright Debate
A composer and campaigner for defending artists’ rights, said the breached music would likely be used for training AI models.
“Training on illegally sourced material is unfortunately widespread in the tech sector, so this accessed music is almost certain to be used instructing AI models,” he said.
The Anna’s Archive site makes references to a large online archive, a huge online repository of pirated books that has reportedly been used by Meta to train its AI models.
A founder of an AI startup wrote on the professional network that members of the public could in theory “create their own personal free equivalent of the service”. He said it could also allow tech companies to “educate models on contemporary music at a large volume”.
Copyright has become a key issue between artists, authors and creatives on one side and AI companies on the other.
Government Proposals and Reactions
Across Britain, creative professionals have raised concerns against a government suggestion to let AI companies use copyright-protected work without permission unless the rights holder opts out. Virtually every respondent to a government consultation on the proposal has supported artists’ concerns.
A government minister for science, research and technology informed parliament this month there was “little consensus” on the matter, adding that ministers would “use the time to get this right”. The government has pledged to make regulatory plans on AI and copyright by March 2025.
