aipopdb

17 Oct 2026  ·  2,818 works · 695 news indexed

News / GEMA launches PLAI, a licensed dataset for training music AI

GEMA launches PLAI, a licensed dataset for training music AI

AIPopDB Newsroom · via digitalmusicnews.com · · 1 min read

Read the original ↗
Image for GEMA launches PLAI, a licensed dataset for training music AI

Image: digitalmusicnews.com

German collecting society GEMA has opened PLAI by GEMA, a training dataset of close to 180,000 fully cleared recordings in more than 60 genres, with detailed metadata, which AI developers can license and adapt.

Five German production-music and publishing houses, among them Sonoton and Intervox, supplied the first catalogue and will be paid in proportion to use. GEMA says other rightsholders can join. The first customer is Klangio, a Karlsruhe company that turns audio into sheet music, which says it wanted training data with no legal doubt attached.

GEMA pitches the set as a way for developers to avoid costly lawsuits. It already won a case against OpenAI, and at the time of writing was awaiting a 31 July ruling in its suit against Suno, which it said should confirm that training must be paid for wherever it happens.