Deezer’s first public transparency report on generative AI reveals that 44% of new music uploads in early 2026 were machine-made, with the majority of associated streams flagged as fraudulent , exposing a systemic attack on how streaming royalties get paid out.
The numbers are stark. Between January and March 2026, nearly one in two tracks uploaded to Deezer was identified as AI-generated. That alone would be a significant industry milestone. But the real story, buried in the French streaming platform’s transparency report released Monday and titled ‘The Artificial Paradox,’ is what’s happening to the money. Deezer’s fraud detection systems , rolled out in late 2025 , found that the majority of streams on that AI content are fake, manufactured by automated bots to mimic real listening behavior. The company estimates this is siphoning tens of millions of euros in royalties each year away from human artists.
This is not a passive overflow problem. Deezer’s report points to organized content farms operating across multiple international jurisdictions, uploading thousands of AI-generated tracks daily with the deliberate intent of gaming the platform’s payment infrastructure. These actors understand exactly how pro-rata royalty models work: the more streams a track accumulates relative to total platform streams, the larger its cut of the monthly royalty pool. Flood the pool with fake engagement, and you quietly redirect revenue at scale.
CEO Jeronimo Folgueira has been direct about what he wants in response: a mandatory tagging system that distinguishes AI-generated audio from human-created work at the point of upload. This isn’t a new idea in principle , the broader music industry has debated AI labeling for years , but Deezer is now attaching hard numbers to the argument, which changes the political weight of the ask. The report stops short of naming specific operators, citing active legal investigations, but the implication is clear: this is coordinated, not coincidental.
The pro-rata model has long drawn criticism from independent artists who argue it inherently favors catalog depth over genuine fan relationships. A superstar with passive background listeners earns more than a niche artist with a devoted but small audience. What Deezer’s report now demonstrates is that the model isn’t just imperfect , it’s actively exploitable by anyone with access to cheap generative audio tools and a bot network. The barrier to running this kind of operation has collapsed in the past two years as both AI music generation and stream manipulation services have become commoditized.
The alternative model that keeps resurfacing in these conversations is user-centric payouts, where each subscriber’s monthly fee flows only to the artists that subscriber actually listened to. It’s been piloted by smaller platforms and studied extensively, but the major streaming services have resisted wholesale adoption. Deezer’s report may not force their hand immediately, but it does make the status quo harder to defend publicly.
For regulators, this is fresh ammunition. The European Union’s AI Act introduced labeling requirements for certain AI-generated content, but music streaming platforms operate in a gray zone where enforcement is still being defined. Deezer is essentially lobbying, through data, for those standards to be tightened and applied upstream , at the distribution layer, before content ever reaches listeners.
What to watch now is whether the other major platforms , Spotify, Apple Music, Amazon Music , follow with their own disclosures, or stay quiet and let Deezer absorb the scrutiny alone. Spotify has acknowledged AI fraud as a growing concern in earnings calls but has not released comparable figures. If Deezer’s 44% figure reflects an industry-wide reality rather than a platform-specific one, the conversation about how recorded music gets monetized in the generative AI era just moved from theoretical to urgent.
Also read: OpenAI’s latest image model just made every competitor rethink their roadmap • An anonymous developer claims to have trained a 235-million parameter LLM on a single consumer GPU and the AI community is not convinced • Apple is betting that the future of AI lives in your pocket not in the cloud
‘ The preceding article may include information circulated by third parties ’
‘ Some details of this article were extracted from the following source startupfortune.com ’
