In a bid to combat the rising tide of streaming fraud, Deezer, a music streaming service based in Paris, announced its plans to start flagging albums containing AI-generated tracks. This initiative comes as the platform grapples with an increase in music created via artificial intelligence tools that are suspected of being used to earn royalties dishonestly.
Deezer will introduce an on-screen notification warning listeners when they’re encountering “AI-generated content.” This move aims to inform users about the presence of tracks produced by song generators within albums. Although Deezer is smaller compared to giants like Spotify, Amazon, and Apple, the company recognizes AI-generated music as an “industry-wide issue.” Deezer is resolute in its commitment to “safeguarding the rights of artists and songwriters” amidst ongoing challenges to copyright law driven by AI model training, as expressed by CEO Alexis Lanternier.
The advent of generative AI systems, which are trained on online text, images, and audio, has significantly disrupted the industry. Consequently, AI companies are now facing multiple lawsuits over claims of scraping web content without due remuneration. Deezer, employing its AI song detection tool, discovered that a striking 18% of daily uploads on its platform—equating to roughly 20,000 tracks—are entirely AI-generated. Just three months prior, this figure stood at 10%.
Lanternier acknowledges AI’s benefits but highlights the “many questions” it poses for the music industry. While AI’s use in music creation is acceptable when an artist is involved, issues arise when anyone, including bots, can use it to produce music. He identified a trend of music fraudsters producing numerous tracks to gain inclusion on playlists and subsequently reap royalties. Yet, although direct uploads by musicians to Deezer or its competitors, such as Spotify and Apple Music, aren’t possible, music labels or digital distribution platforms can do so for contracted artists, and independent users have self-service options.
Currently, purely AI-generated tracks represent a scant 0.5% of total streams on Deezer. Nevertheless, the platform perceives “fraud” as the primary intent behind these songs, suspecting that up to 70% of streams are executed by “farms” or bots, not actual listeners. Deezer plans to withhold royalty payments from any AI songs identified as part of “stream manipulation.”
The subject of AI has stirred substantial debate in the music industry, oscillating between its creative potential and the scrutiny of its legality. Notably, AI song generators like Suno and Udio are embroiled in copyright infringement lawsuits, accused by record companies of unlawfully using recordings from artists like Chuck Berry and Mariah Carey. Similarly, Gema, a German royalty collection group, is suing Suno over allegations of creating songs strikingly similar to originals by artists it represents, including hits like “Forever Young” by Alphaville and Boney M’s “Daddy Cool.”
Major record labels are reportedly negotiating compensation with Suno and Udio, according to recent reports. Deezer uses song generators themselves to analyze content for tagging, utilizing complex patterns discernible within AI-produced music. According to Lanternier, keeping detection tools current is vital given that the recognizable patterns in AI music are continuously evolving, leading Deezer to continually generate AI songs for algorithm training. “We’re fighting AI with AI,” he remarked.
The profitability of streaming fraud is significant. Lanternier references a notable U.S. criminal case last year—the first involving inflated music streaming—where prosecutors charged a man with wire fraud conspiracy. The accused had generated hundreds of thousands of AI songs, used bots for automatic streaming billions of times, and purportedly earned at least $10 million.