Universal Music Group says AI-generated tracks are arriving on streaming services at a rate of more than 90,000 a day, and argues that labelling, licensing and payment rules need to catch up. Its proposed answer is not to ban AI, but to stop undisclosed or unauthorised material from borrowing the value of artists’ work without consent or fair payment.
Watch Desk analysis
What happened
In an interview with Diginomica, UMG chief executive Lucian Grainge said the company estimated roughly 10,000 AI-generated tracks were arriving daily in early 2025, compared with more than 90,000 now. UMG also says that, on some digital music services’ Blues charts, as many as five of the top ten albums are by AI-generated acts without clear labelling. These are company claims, not an independently audited count.
UMG’s chief digital officer, Michael Nash, described four protections the label seeks in agreements with streaming services: no training on its artists’ work without consent; removal of unauthorised AI recordings trained on that material; removal of content that misuses artists’ identities; and safeguards intended to prevent fully AI-generated music diluting artists’ royalties. Nash also pointed to Spotify removing more than 50 million tracks from recommendations and 75 million spammy AI-generated tracks from its service, and to Deezer’s planned option to exclude fully AI-generated music.
Why it matters
The dispute is about more than whether a track was made by a person or a model. It touches disclosure, training permission, impersonation and how streaming revenue is divided. If listeners cannot tell what they are hearing, and payment systems treat every stream alike, the rules may reward volume over the artists and music that bring people to a service.
The figures and proposed protections come from UMG and its executives, who have a clear commercial stake in how streaming agreements are written. They make a substantial case for scrutiny, not a settled measure of the size or impact of the problem.
Our read
UMG is right to put practical questions on the table: what was used to train a system, who authorised the result, how is synthetic material labelled, and who gets paid? The eyebrow-raising part is how much depends on private platform agreements. “AI music” is not one tidy category, and a useful policy needs to distinguish licensed experimentation from impersonation and spam, rather than asking listeners to solve the mystery themselves.
What to watch
- Whether streaming services publish clearer rules for labelling and recommending AI-generated music.
- Whether UMG’s proposed protections appear in more platform agreements, and how they are enforced.
- Whether payment models distinguish licensed AI work, unauthorised imitation and human-led recordings.
Discussion spark: Should fully AI-generated tracks earn the same streaming royalties as human-led music, or should payment depend on how the music was made and licensed?
Sources and evidence
- Want to hear more about Generative AI and LLMs? (16 November 2026, 12:19 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.