After years of letting algorithms learn for free, record labels now want their cut. AI companies have quietly trained systems on massive music catalogs without paying — but now, Universal MusicUNVGY

and Warner MusicWMG

are negotiating landmark licensing deals to change that. Rather than go to court, the industry is racing to lock in agreements before AI fully reshapes how music gets made.

Sending the bots a bill: At the center of negotiations is a streaming-style model where AI companies would pay micropayments each time they use licensed music — similar to how SpotifySPOT

pays royalties. Labels also want attribution tools like YouTube’s Content ID system under GoogleGOOGL

, which can detect when their catalogs are tapped for training or generating new tracks. Atlantic Records CEO Elliot Grainge says he’s “very bullish” about the opportunity, though whether artists embrace the deals remains an open question (FT).

• The major labels are in active talks with AI startups like Suno, Udio, ElevenLabs, and Stability AI, as well as tech giants such as Google and Spotify.
• Some of the negotiations may also involve settling past unauthorized use, with labels hoping to establish legal ground rules before AI music becomes impossible to control.

Drowning Out The AI Noise

The urgency makes sense when you see what’s already flooding streaming platforms. Spotify recently removed 75M “spammy” tracks, with much of it being AI-generated music that’s cutting into royalty pools. French streaming service Deezer says nearly a third of daily uploads are machine-made, with roughly 150K tracks hitting platforms every day — a volume three orders of magnitude greater than anything the CD era produced. Artists worry their music is being pulled into training systems without credit or pay, while labels juggle lawsuits against the same companies they’re also striking deals with.

• Deezer estimates 70% of fully AI-generated tracks involve fraud — from fake artist profiles or schemes designed to siphon royalties from legitimate musicians.
• To tackle some part of this, Spotify is rolling out new policies against impersonation and spam, along with filters to block mass uploads and duplicates enabled by AI.

Learning from past mistakes: Back in 1999, when Napster disrupted the industry, labels spent years in court while their business collapsed — losing more than half their value before finally conceding that digital music was here to stay. That lesson is shaping their response today. Suing every new tool isn’t realistic, so labels are choosing to partner with AI firms in an effort to get ahead of the technology. The challenge now is making sure music doesn’t just feed the machine — but also pays the people who create it.