Deezer Has Found 42 Million AI-Generated Tracks. The Music Industry Is Done Waiting.

Deezer has identified over 42 million AI-generated music tracks on its platform — a number that reframes the scale of the problem streaming services have been quietly managing for the past two years. The disclosure arrives as major music publishers escalate legal and lobbying campaigns against AI companies, and as the industry prepares for a public reckoning at the AI Creative Summit in London on November 11.

How It Got to 42 Million

The AI music generation market did not exist at commercial scale two years ago. Today, tools like Suno, Udio, and a dozen lesser-known platforms allow anyone with a prompt to produce a passable song in seconds — and upload it to streaming services at scale. Some operators have built fully automated pipelines: generate tracks, create artist pseudonyms, upload, collect micro-royalties from algorithmic playlist placements.

The scale of that activity is why Deezer's 42 million figure is significant. Deezer has roughly 120 million total tracks in its catalog. If those numbers are accurate, more than one-in-three tracks on Deezer is AI-generated. The ratio almost certainly reflects the upload velocity of automated operators more than it reflects the share of what people actually listen to — but it creates a catalog hygiene and royalty distribution problem that the streaming services can no longer treat as marginal.

Deezer has responded with a three-pronged approach: AI detection and labeling, removing tracks from royalty-eligible pools, and filtering fully AI-generated content from human-curation surfaces. Spotify and Apple Music have implemented similar but less publicly detailed systems.

What Music Publishers Are Fighting Over

The streaming services' internal policies are addressing the distribution problem. The publishers' legal campaigns are targeting the upstream cause: the training data.

Major music publishers — including Universal Music Group, Sony Music, and Warner Music Group — have filed and expanded lawsuits against AI music generation companies, arguing that models like Suno and Udio were trained on copyrighted recordings and compositions without consent or compensation. The publishers allege this constitutes both copyright infringement and, in some cases, a deliberate attempt to launder the training data problem by describing it as "inspiration" rather than reproduction.

The most recent escalation, documented by Further & Better, involves publishers lobbying governments in the US, UK, and EU to reject proposed copyright exceptions that would allow AI training on copyrighted material without a license. The publishers argue AI companies are not just copying their content — they are simultaneously lobbying to make that copying legal retroactively.

That framing — and the 42 million track figure — will be front and center at the AI Creative Summit at BFI Southbank in London on November 11, which has positioned itself as the event where the creative industries formally present their unified position on AI and copyright to policymakers.

The Royalty Distortion Problem

Beyond the legal fight, there is a structural economics problem that AI-generated tracks create for human musicians.

Streaming royalties are distributed based on a pro-rata model: each stream earns a fractional share of the total royalty pool. When tens of millions of AI-generated tracks capture even small numbers of streams — through playlist placements, background listening, or sleep/ambient playlists — they dilute the royalty pool available to human artists without contributing meaningfully to the listening experience users pay for.

Some streaming services have begun moving toward a "fan-powered" royalty model that allocates royalties based on what individual subscribers actually listen to, rather than pooling all streams platform-wide. This reduces the dilution effect of AI tracks that no single subscriber deliberately chooses. But the transition is slow, and most royalties still flow through the traditional pro-rata system.

Where the Legal Cases Stand

Suno has filed a formal legal response to the publisher lawsuits, contesting the copyright infringement claims on fair-use grounds and arguing that generative AI trained on music creates something new rather than copying what it learned from. The legal theory is contested — courts have not yet reached a final judgment in any of the major AI-music copyright cases — but Suno's willingness to fight rather than settle signals it believes the fair-use argument is viable.

The outcome of these cases will set precedents that extend well beyond music. The same training-data questions apply to AI image generation, AI writing, and AI video. A ruling that AI training on copyrighted material without a license is infringement would reshape the economics of every major AI foundation model company.

What to Watch

The November 11 AI Creative Summit will give publishers, labels, and musician rights organizations a public platform to present their case for legislative action before year-end. In the US, the Senate Judiciary Committee's IP subcommittee has been holding AI-and-copyright hearings throughout 2026; a bill is reportedly in draft. In the UK, the government's AI and Creative Industries working group is expected to publish its recommendations before the end of 2026. In both cases, the 42 million tracks figure gives advocates a concrete data point that is harder to dismiss than abstract legal arguments about training-data consent.

By Hector Herrera