AI audio company ElevenLabs has launched Eleven Music, a text-to-music generator, by striking preemptive licensing deals with independent music licensor Merlin and publisher Kobalt. The move is a direct attempt to avoid the massive copyright lawsuits currently plaguing competitors like Suno and Udio.
- Playing by the book: While other AI music startups face legal battles with the RIAA over alleged mass copyright infringement, ElevenLabs is building its platform on a foundation of consent. The deals allow artists represented by Merlin and Kobalt to opt-in, licensing their work to train a more advanced "Pro" model in exchange for a cut of the revenue.
- Upending the economics: The deal with Kobalt, first reported by Music Business Worldwide, establishes a new precedent for AI royalties, delivering an approximate 50/50 "parity" split between the song's publisher and the sound recording's owner. This is a clean break from lopsided streaming economics and is protected by a clause ensuring Kobalt's terms will match any better deal negotiated in the future.
With a framework in place for the independent sector, ElevenLabs has its sights set on the industry's biggest players. In a Bloomberg Live interview, CEO Mati Staniszewski confirmed his ambition to eventually partner with major labels like UMG, Warner, and Sony.
-
The new blueprint: ElevenLabs is betting that collaboration, not confrontation, is the path forward for generative AI in music, creating a potential playbook for an industry wary of unlicensed training data.
-
Also on our radar: The launch comes as AI's impact on music becomes more tangible, with one fully-AI band racking up millions of listens on Spotify, raising new questions about disclosure. Meanwhile, ElevenLabs' powerful voice-cloning tools have a history of controversial use, including being deployed to create a deepfake Joe Biden for political robocalls, underscoring the ethical tightrope these technologies walk. And this isn't the company's first licensing rodeo; it previously inked a deal with sync platform SourceAudio to train its models on legally-cleared music.