AI just got a record deal
Spotify struck a licensing deal with Universal Music for fan-made AI covers and remixes, launched Studio as an AI podcast generator, and shipped an ElevenLabs-powered audiobook tool — while Stability AI released open-weight models capable of generating six-minute songs. The strategy is shifting from resistance to revenue.
TechCrunch
Spotify and Universal Music strike deal allowing fan-made AI covers and remixes
Spotify and Universal Music strike deal allowing fan-made AI covers and remixes.
techcrunch.com

This week, Universal Music Group signed a licensing deal with Spotify allowing fans to create AI-generated covers and remixes. A record label choosing to monetise fan creativity rather than police it. If you know the music industry, you know this was always coming.
The oldest play in the book
The pattern is familiar: a new technology threatens existing revenue, the industry pushes back, and eventually someone figures out how to put a price tag on it. AI appears to have reached the licensing phase.
The evidence arrived in a single day. Beyond the Universal deal, Spotify launched Studio, an AI podcast generator rivalling Google's NotebookLM. Hours later, the company shipped an ElevenLabs-powered audiobook tool for AI-narrated audiobooks.
Three AI audio products from a company built around listening, not creating. That's not a change of heart. That's a company that figured out where to put the meter.
Meanwhile, Stability AI released Stable Audio 3.0 with open-weight models for six-minute songs. The capability is open, and each release pushes the boundaries further.
Who holds the lever
The interesting question isn't whether AI-generated audio is acceptable now. It's who controls the terms. The Spotify-Universal deal channels fan creativity through a licensed framework where the label gets a cut. This is the record industry doing what it has always done: building toll booths on other people's creativity. Fan remixes existed long before generative models. What's new is that Universal found a way to monetise them without having to produce, approve, or even hear each one.
For builders, the lesson is structural. If your AI product touches copyrighted material, the question is no longer whether you'll need a licence but what percentage you'll surrender. Audio may have moved first because its rights infrastructure was comparatively well-established. Text and image are further behind, but the template is now set.
The one variable is open-weight models. Stability's decision to release open weights creates a parallel track where the toll booth is optional. If a bedroom producer can generate a six-minute track locally with no platform in the loop, the licensing framework has a leak.
The music industry has learned, repeatedly, that you cannot sue a technology out of existence. The new bet is that you can price it instead. Whether that bet holds depends on how long the open models stay niche.
Read the original on TechCrunch
techcrunch.com