{"id":106088,"date":"2025-09-30T09:48:05","date_gmt":"2025-09-30T08:48:05","guid":{"rendered":"https:\/\/routenote.com\/blog\/?p=106088"},"modified":"2025-09-30T09:48:07","modified_gmt":"2025-09-30T08:48:07","slug":"can-we-finally-detect-copyrighted-works-in-ai-music-umg-and-sony-join-forces-with-soundpatrol","status":"publish","type":"post","link":"https:\/\/routenote.com\/blog\/can-we-finally-detect-copyrighted-works-in-ai-music-umg-and-sony-join-forces-with-soundpatrol\/","title":{"rendered":"Can we finally detect copyrighted works in AI music? UMG and Sony join forces with SoundPatrol"},"content":{"rendered":"\n

New audio fingerprinting tech could change how the industry tackles copyright in AI-generated music.<\/span><\/h2>\n\n\n\n

A big step for copyright in the AI era<\/h3>\n\n\n\n

Universal Music Group and Sony Music have teamed up with SoundPatrol<\/a> on a new tool designed to spot copyright infringement in music, including tracks partly or fully generated by AI. The technology, described as the \u201cfirst-of-its-kind\u201d, aims to protect creators and their intellectual property.\u00a0<\/p>\n\n\n\n

For a music industry grappling with the rise of AI music generators like Suno<\/a>, this could be a turning point. Artists have long worried about their work being scraped without permission to train AI models, only for their style and sound to pop up in AI-generated tracks. SoundPatrol\u2019s system is built to detect copyright infringement, keep artists in control, and hopefully the revenue too.<\/p>\n\n\n\n

\u201cWe\u2019re constantly focused on enabling AI- bringing to market the many commercial and creative opportunities that will benefit our artists while establishing effective tools to protect them. Bringing solutions to the table that support the entire industry\u2026\u201d<\/strong><\/p>Sir Lucian Grainge, UMG\u2019s Chairman and CEO<\/cite><\/blockquote>\n\n\n\n

How does the tool actually work?<\/h3>\n\n\n\n

At its core, SoundPatrol\u2019s system uses what it calls a \u201cforensic AI model for audio-video fingerprinting\u201d. Rather than relying on traditional methods that match exact audio, this tool goes deeper. It uses neural embeddings- essentially the DNA of a track. By looking at the semantics of a piece, it can detect whether human-made content shows up in AI output, and even includes remixes, covers, or derivatives. <\/p>\n\n\n\n

In fact, it appears to work in a similar way to a tool that the Swedish collecting society STIM are working on<\/a> for effective AI music licenses.<\/p>\n\n\n\n

\n
Collecting society, STIM launches pioneering AI music licence for creators<\/a><\/blockquote>