German collecting society GEMA has launched PLAI, a copyright-cleared music dataset intended to give AI developers a single licensed source for training data. Its launch matters because it offers an alternative to disputed use of copyrighted music in AI development, while GEMA awaits a court decision in its case against music generator Suno.

PLAI combines sound files, metadata, composition rights and master rights, and is designed for creator-assistance tools whose outputs do not compete with their training material. Its first customer, transcription firm Klangio, will use a dataset of about 178,000 sound files spanning more than 60 genres; GEMA says clients can tailor the data and additional rights holders may join. GEMA sued Suno in January 2025 over alleged unlicensed training on copyrighted recordings, with a ruling expected on 31 July.