In a November 9 update, Sound Ethics said researchers at UC Berkeley and New York University are testing how reliably different tools can identify songs that influenced an AI-generated track. The teams are running the methods side by side under the same conditions to see which works they identify, where they disagree and how clearly they can explain a match. The testing is set to continue through July 2027.
Sound Ethics, an independent research organization based in Hollywood, California, works with university labs to study how AI systems use existing music and develop tools that help artists and rights holders identify when their work was used.
The initiative also looks at whether researchers can identify specific songs in a model’s training data. For publishers and labels, that answers a different question from tracing the influences on a single track. One could help them check whether an AI developer used works from their catalog. The other could help them assess whether a particular output drew on those works. If the methods prove reliable, rights holders could use that evidence to negotiate what a license covers and how payments are calculated.


Join the discussion