Emails: UMG asked Spotify, Apple, and other services to block developers training AI services from scraping melodies and lyrics from their copyrighted songs
Universal Music Group says new technology relies on unauthorised use of copyrighted material — Universal Music Group …
Context & Ripple Effects
UMG's request to Spotify, Apple and other services is the opening move in the labels' campaign over AI training data: rather than suing first, the world's largest music company tried to cut off scraping at the distribution layer by enlisting the platforms that host the catalogs. The strategy then escalated on two tracks — legal, with UMPG, Concord and ABKCO filing a copyright suit against Anthropic over song lyrics months later, and commercial, with Google and UMG entering licensing talks for melodies and voices.
By 2024 the pattern had gone industry-wide, as Sony sent warning letters to more than 700 AI companies, developers and streaming platforms over unauthorized use of its content — confirming that UMG's platform-level pressure was a template, not a one-off.
First-order effects
- Spotify, Apple and other services are being pushed into an enforcement role they didn't ask for: policing developer access to melodies and lyrics, or facing label pressure over hosting infringing pipelines.
- AI developers building music models lose frictionless access to the most valuable training corpus in music — major-label catalogs — unless they negotiate licenses.
Second-order effects
- Labels convert enforcement leverage into deal-making power: the same companies blocking scrapers open licensing negotiations, as seen in the Google–UMG talks, turning training data from a free input into a priced product.
- Competing rights holders copy the playbook — Sony's mass warning letter campaign mirrors UMG's approach, standardizing 'warn, then litigate or license' across the industry.
Third-order effects
- If platforms accept the gatekeeping role, music AI develops inside licensed walled gardens, with major labels controlling which models can train at all — and unlicensed AI music facing downweighting or bans on streaming services, as later commentary around Spotify has urged.
The trend: Music rights holders are moving from passive complaint to active control of the AI supply chain, using platforms, lawsuits and licensing talks to make label consent a prerequisite for training on songs.