Streaming giant Spotify has deleted over 75 million tracks as it comes under fire from an industrial-scale wave of artificial intelligence-generated music.
As AI giants make it increasingly easier to make fake music, it is piling huge pressure on services like Spotify and the systems they use to pay real artists.
The proof is already here in Australia, as a suspected AI-generated song tops the charts while artists complain they have to share a stage with AI-assisted artists.
Sam Duboff, the key executive in charge of global artists, marketing and policy at Spotify, said 100,000 songs were uploaded to the platform’s servers every day and an increasing number of them were generated by AI.
Speaking to the AFR as he visited Australia this week for talks with artists, labels and policymakers about the impact of AI, Mr Duboff said Spotify now had an entire team dedicated to detecting AI music.
“That’s the low-effort content you see where it’s mass-produced, it’s mass uploaded, it’s AI slop … we’ve removed over 75 million spammy tracks in the last year,” he said.
“AI hasn’t really introduced any novel spam tactics. It’s just taken existing spam tactics and taken them to a new level.”
He added it was a difficult situation for the platform as there were blurred lines about AI use in music.
Spotify is being hit from all sides as real artists are increasingly adopting AI in the studio, everyday creators are scoring viral TikTok and Instagram hits with AI-fuelled remixes, and deepfakers are churning out clone bands and vocal tracks.
Making matters worse, AI giants like OpenAI and Anthropic continue to aggressively scrape online data to train their models.
“We have built lots of systems to try to make scraping as difficult as possible,” Mr Duboff said.
“We certainly have a big team trying to look at any sort of attack vector that might be coming in. So, you know, we do our best to protect all the music that people trust us with.”
‘Clear message’: Artists back Albo’s AI plan
The Spotify executive’s Australia visit comes as Anthony Albanese announced a big move to combat the rise of AI — revealing he would create an Office of Artificial Intelligence.
The peak body representing some of Australia’s biggest artists has backed the plan, hailing it as a “clear message to AI companies”.
The Prime Minister on Wednesday used a speech in Sydney to tout the new office, nestled within the Department of Prime Minister and Cabinet, as a world-leading approach to AI.
“Getting this right will enhance our appeal to international investors by delivering greater clarity and speed for approvals and a streamlined process for verifying compliance,” Mr Albanese said.
“It also imposes an important discipline on government.
“AI touches on the work of every minister and department, so it is only natural that, up until now, our response has been issue by issue, sector by sector.”
The Australian Recording Industry Association (ARIA) chief executive Annabelle Herd welcomed the federal government’s move.
“The Prime Minister could not have been clearer: Australian writers and musicians keep ownership and control of their work,” she said.
“Artists control what that work is worth, not the government and not a technology company.
“Control of price, value and terms of use are what underpin a commercial licensing market. The artist decides what their work is worth and who may use it.
“That is how licensing works everywhere else in the world and it is how it should work here. In the Prime Minister’s words: anything less is theft.”
It comes just two weeks after some of Australia’s biggest creatives pleaded with Labor to stand firm on pledges not to tweak copyright laws.
The group, which included singer-songwriter Jack River, called on AI companies to enter into licensing agreements.
“This is a clear message to AI companies: now is the time to get on with licensing,” Ms Herd said.
“Right now deals are being signed across music, journalism and publishing around the world.
“Australia’s creative industries are ready to do business.”
Music industry launches AI-generated content labels
The move comes after several major music industry organisations unveiled a labelling system for content created with generative AI that they would like to see widely adopted.
The International Federation of the Phonographic Industry (IFPI) and the Recording Industry Association of America (RIAA) announced the voluntary labels alongside six other groups including the Grammys.
“Fans want to know whether and how generative AI has been used,” the chief executives of IFPI and RIAA said in a prepared statement.
“These labels will provide an immediately understandable and easily scalable approach to transparency.”
They unveiled two labels. The first would indicate music that is primarily “AI-generated” — cases where artificial intelligence “was used to generate the entirety or the primary portion of the creative elements of the recording”.
This includes tracks generated “entirely” from AI prompts, as well as lead vocals and “key” instrumental tracks that are AI-generated, according to the statement.
The second label applies to “AI-assisted” music recording which are still “created substantially by humans and expresses human creativity” but contain “some expressive elements” that were generated with AI.
However, humans must perform the lead vocals and primary instrumental tracks. This voluntary labelling system is designed for “broad, global adoption,” including on streaming services.
Music streaming site Deezer systematically flags tracks generated with AI, which the company recently said appear in close to half of new uploads. In June, it launched an “AI music detector” which it said is 99.8 per cent accurate.
Earlier this year, an Apple Music executive told Billboard more than one third of new uploads were entirely created with AI.
The Digital Media Association, a trade group representing streaming companies including Apple Music, Amazon and Spotify, said it was following the labelling announcement closely and looks forward to receiving more detailed and accurate AI metadata as a way to “strengthen our ability to give fans the transparency they deserve”.
“DIMA has long advocated for the creators, owners, and distributors of music to provide accurate and timely metadata on all music released and distributed to streaming services,” the association’s CEO Graham Davies said in a statement.
In April, Spotify launched a “Verified by Spotify” label to signal that users can “trust the authenticity” of an artist, and last year the company announced new efforts to support AI disclosure and combat impersonation.
— with NewsWire and AFP