Sweden’s performing rights organization STIM has signed what is believed to be the first-ever licensing agreement between a music rights society and an AI music generator, marking a major milestone in how artificial intelligence is treated within the legal frameworks of music creation and compensation.
The deal was struck with SonicLoom, a Stockholm-based AI music startup that enables users to generate fully produced compositions using natural language prompts. Under the terms of the agreement, SonicLoom will license musical works from STIM’s repertoire and implement a royalty-tracking system that compensates rights holders whenever the AI-generated outputs are based on or trained from registered compositions.
According to STIM, the move is designed to ensure that songwriters and composers are protected and paid in an era where creative input is increasingly supplemented—or replaced—by machines. The organization emphasized that this agreement does not legalize unlicensed training on copyrighted music, but rather provides a framework for regulated, transparent use of intellectual property in AI development.
The agreement addresses a core tension at the heart of the AI-music debate: where does inspiration end and infringement begin? While many AI platforms claim to generate “original” material, critics argue that training algorithms on copyrighted catalogs without permission amounts to mass-scale sampling without credit or compensation. By proactively licensing use of its repertoire, STIM is attempting to bring AI creators into a more accountable relationship with the rights-holding ecosystem.
SonicLoom will also implement a music identification and fingerprinting system to detect when AI outputs closely resemble real-world compositions. When applicable, partial royalties or usage credits will be triggered and distributed through STIM’s rights management infrastructure.
For music publishers and rights holders across Europe, this development sets a precedent that could ripple into new licensing models for AI across genres, markets, and platforms. It also challenges major tech companies and AI startups—some of whom have trained on copyrighted material without consent—to begin negotiating with collecting societies rather than operating in legal gray zones.
On a global scale, the deal is already drawing attention. U.S.-based rights groups and publishers have largely taken a more combative approach to AI music—filing lawsuits, issuing takedowns, and lobbying for government intervention. Sweden’s more collaborative, infrastructure-driven approach presents an alternative: create the rules instead of just enforcing them.
Whether this licensing model becomes a template or a one-off will depend on adoption, enforcement, and transparency. But one thing is clear: the music industry is no longer asking if AI will be part of its future. The question now is who controls it—and who gets paid.
