Sony Tech Can Identify Original Music in AI-Generated Songs

Sony Group has developed technology that can identify underlying music used in songs generated by artificial intelligence (AI), potentially enabling songwriters to seek compensation if their work was used, Rei Nakafuji reported for Nikkei News.

Sony believes the technology could help create a revenue-sharing system that compensates original songwriters based on their contributions. (Photo: Kyodo News) 

The technology analyzes which musicians’ songs were used in training and generating AI music.

It can quantify each original work’s contribution, for example estimating that a track contains “30% Beatles and 10% Queen.”

If an AI developer cooperates, Sony can obtain data by connecting directly to the developer’s base model system. If cooperation is not possible, the system estimates original influences by comparing AI-generated music with existing works.

The AI boom has led to numerous allegations that developers used copyrighted music, video, and writing without permission to train models. AI-generated songs using the voices of well-known singers have also circulated online.

Sony believes the technology could help create a revenue-sharing system that compensates original songwriters based on their contributions.

In Japan, copyright law protects two categories of music rights: copyrights held by songwriters, composers, and publishers, and neighboring rights held by performers and record producers.