GEMA Launches Fully Cleared ‘PLAI’ Music Dataset for AI Training

An aerial shot of Karlsruhe, Germany, where Klangio, the first PLAI by GEMA client, is based. Photo Credit: Carsten Steger

Nearly two years after announcing plans to develop a gen AI licensing framework, GEMA has launched PLAI, billed as today’s “first complete music dataset for simple AI training.”

The Berlin-based society reached out to DMN with word of “PLAI by GEMA” as well as the “curated and individually adaptable” dataset’s first AI client. On the former front, PLAI is said to encompass almost 180,000 recordings spanning north of 60 genres.

And while it probably doesn’t need saying, this library is fully cleared and comes equipped with “comprehensive metadata,” per GEMA.

Regarding the initial music-world participants, Germany’s Bailer Music Publishing, Earmotion Audio Creation, Intervox Production Music, Music Sculptor, and Sonoton Music are aboard. They’ll be compensated “proportionately and appropriately,” the society stated.

Moreover, different “publishers, collecting societies, authors and rights providers can be integrated into the growing ecosystem at any time,” according to GEMA. That’s important generally and because the entity is litigating against AI giants including Suno.

As we noted last week, a decision in that case is expected on July 31st, and GEMA previously emerged victorious in a courtroom confrontation with OpenAI.

In describing PLAI’s “main advantages” for clients, GEMA emphasized one-stop licensing, “flexible data packages” (tailorable by genre, mood, and more), and guaranteed compliance, to name a few.

Building on the latter perk, the opportunity to avoid costly legal battles and headaches jumps out as a key selling point. Back to the customer side, then, AI-powered music transcription business Klangio has signed on as the first to use PLAI, for which companies can request quotes via GEMA’s website.

“Our customers and we ourselves are part of the music industry,” Klangio head Sebastian Murgul said. “That is why fairness and legal certainty are a major issue for us. With the GEMA music dataset, we have found the solution we have been waiting for: legally fully clarified training data where all rights are really covered.

“This gives us the security to further develop our AI tools on a solid foundation. This shows that fair remuneration and innovative AI development can go hand in hand,” he concluded.

And in remarks of his own, GEMA director of broadcasting and online Thomas Theune touted PLAI’s role in “making music content for training AI tools more accessible than ever before.

“Instead of complex rights clearances,” Theune continued, “AI tool providers now receive high-quality music data including the necessary rights from a single source. This creates legal certainty for developers and new revenue potential for music creators.

“With PLAI by GEMA, we show that AI training and fair remuneration of creatives are not a contradiction. If both sides are interested in dealing responsibly with each other, we will find joint solutions and enable innovation on a reliable legal basis,” the longtime GEMA higher-up finished.

Returning to the Suno litigation, GEMA general counsel Kai Welp underscored the case’s broader AI licensing significance and the need for a definitive “political framework.”

“For the licensing of generative AI providers,” Welp added in part, “we hope for a positive decision on the matter against SUNO on July 31. It must be made clear that the use of creative works must be remunerated regardless of the location of the training and also when the original works are stored in the systems. This requires the courts, but above all the political framework.”