GEMA sues for fair compensation

GEMA is the first collecting society worldwide to file a lawsuit against a provider of generative artificial intelligence (AI) systems for the unlicensed use of protected musical works. Specifically, it concerns the US company OpenAI, the autogenerative chatbot systems operator.

GEMA accuses OpenAI of reproducing protected song lyrics by German authors without having acquired licenses or paid the authors in question. The aim is to prove that OpenAI systematically uses GEMA's repertoire to train its systems.

Go to press release (published on 13 November)

"The new technology presents us with fundamental legal questions that we absolutely must clarify. This is the only way we will succeed in establishing a licensing model on the market that strikes a fair balance between creators‘ and exploiters‘ interests. Our model procedure makes a decisive contribution to this. However, it also shows that we are prepared to enforce the rights to which authors are entitled."

Dr. Kai Welp, General Counsel of GEMA

"Our members‘ songs are not free raw material for generative AI systems providers‘ business models. Anyone who wants to use these songs must acquire a license and remunerate the authors fairly. We have developed a license model for this. We are taking and will always take legal action against unlicensed use."

Dr. Tobias Holzmüller, CEO of GEMA

"Last week, we made it clear with GEMA's AI Charter that human creative achievements must not be used as a free template for the offerings of AI providers in a deeply commercial exploitation chain. Likewise, we cannot accept that copyright infringements occur in the output of chatbots. GEMA's lawsuit sends an important signal: the livelihood of us creative professionals is at stake."

Dr. Ralf Weigand, Supervisory Board Chairman of GEMA

0 / 3

FAQ on the AI lawsuit

Is it correct that GEMA has filed a lawsuit against providers of AI systems?

Yes. GEMA has filed a lawsuit at the Munich Regional Court against OpenAI Ireland Ltd, operator of the AI tool ChatGPT in Europe, and its parent company OpenAI LLC. GEMA will also consider whether to initiate legal action against other AI providers in the future. The aim of these lawsuits is to obtain a licence fee for the authors and music publishers on whose works the AI systems are based. The success of AI systems is based on the content with which they have been trained. It is therefore only fair that those who have created this content should receive a share of the revenues generated by AI systems.

Why is GEMA taking legal action against Open AI?

Technological developments in the AI sector have opened up completely new possibilities for generating music. GEMA is generally in favour of this. It sees AI systems as an important tool in the creative process. However, if content by its members is used to train the systems, they must be fairly remunerated. GEMA was able to establish that song texts from works (lyrics) whose rights it administers are made available to the general public via ChatGPT. This is only possible because this tool was previously trained with these lyrics. Neither the AI’s training with nor the reproduction of the lyrics has ever been paid for by OpenAI. Making protected works available to the public is already subject to a licence under current law. GEMA aims, via the current lawsuit, to enforce this position in favour of the creators against OpenAI. This is not intended to generally prevent the use of GEMA's works by AI systems. The lawsuit’s aim is solely to ensure authors receive fair remuneration when their works are used by AI tools.

Do the providers of AI models and systems have to pay remuneration to the authors for the training?

Yes, this is provided for by law. Legislation does authorise text and data mining, i.e. the systematic internet searching to obtain information about patterns, trends and correlations. However, whether the training of AI models and systems is permitted on this basis alone is highly controversial. In addition, the authors have the option of so-called opting out. This means they can declare that their works may only be used to train AI systems under licence. GEMA has declared this opt-out on behalf of its members. This is intended to ensure that authors receive an appropriate share of the income generated by AI systems. The systems can only produce music recordings because they have been trained with protected works. GEMA members earn their living by licensing their work. GEMA wants to prevent the big tech giants alone from benefiting from the new technology. It is to be feared that revenue generated by their works, which is how these authors earn their living, will in future bypass them and flow to these large companies’ shareholders. This would remove the incentive to be creative. However, human creation is still necessary in the age of AI. AI systems do not work creatively. Their output is based on probability calculations. They have to learn the necessary parameters from human-created works.

May GEMA declare the opt-out on text and data mining for its members and does GEMA have the right to licence AI systems training?

Yes, GEMA had these rights transferred to it by its members very early on. In accordance with this, the GEMA Members' Assembly in 2022 amended the authorisation agreement, i.e. the contract with which the members grant GEMA their rights of exploitation. They also expressly granted GEMA the authority to declare the opt-out. At that time (May 2022), generative AI systems did not yet have the significance they have today, but the members also transferred the rights to so-called unknown use types to GEMA. This ensures that GEMA can comprehensively represent its members at all times even when new technologies emerge. GEMA therefore holds the rights to use musical works for the purposes of generative AI.

Was ChatGPT trained with works to which GEMA holds the rights?

Yes, OpenAI has used protected song lyrics for the training of ChatGPT, the rights to which are represented by GEMA. This can be clearly proven. ChatGPT can play the original lyrics of GEMA works without access to the Internet. The system must therefore have been trained with the original texts. It can also be assumed that the protected texts are still stored in the AI providers’ systems. This also requires a licence under current copyright law.

Does GEMA administer the rights to song texts (lyrics)?

Yes and, in addition, for some years now, GEMA members have also had the option of granting GEMA so-called “graphic rights”, i.e. the rights to the written text image. For example, GEMA not only grants rights to stream music but to simultaneously display the lyrics to major music services via its subsidiaries All authors and music publishers involved in the lawsuit have also given GEMA their separate consent to invoke the graphic rights in the lawsuit.

Why is GEMA taking action against song texts (lyrics) of all things?

In addition to the music played, GEMA also manages rights to song lyrics presented on the internet. However, the lawsuit is also intended to clarify numerous legal problems associated with the use of AI in general. Providers of AI models and systems are currently exploiting these unresolved legal issues to evade the licence fee owed for the use of protected content. The “lyrics lawsuit” is also representative of other uses such as the generation of sound files by unlicensed services. The associated legal problems are similar. Lyrics have the advantage for a lawsuit that infringements can be clearly established. This is much more difficult with music recordings as there is greater room for interpretation regarding whether a composition constitutes plagiarism. Basing the action on lyrics avoids time-consuming evidence-taking in court proceedings, which can be used by the other side to delay court decisions.

Which copyright infringements are the subject of the Lyrics lawsuit?

Of course not every AI platform commits copyright infringements. However, the infringements that GEMA was able to identify with ChatGPT are quite diverse. In addition to use of the original texts without permission, unauthorised adaptations (“hallucinations”) and infringements of moral rights were found. GEMA was also able to prove that Open AI uses original GEMA repertoire to train and further develop its models and systems. The question of whether the copyright barrier covers use of such material to train generative AI systems via text and data mining is already highly controversial. But even if this is permitted, rights holders have the option of preventing the use of their works for AI training by declaring reservation of use. GEMA has declared this reservation of use for its members. The training is therefore illegal in any case.

What does GEMA hope to gain from this lawsuit?

GEMA is hoping to guarantee fair participation for authors in the use of their copyright-protected works by generative AI. The lawsuit is a test case to clarify numerous legal issues. The aim is to specifically refute the AI system providers’ contention that training with and subsequent use of the generated content is free of charge and possible without the rights holders’ authorisation. GEMA wants to establish a licence model on the market in which systems training using copyrighted content, generation of output based on that and the further use of the output must be licensed. In addition, the lawsuit’s filing and the resulting media attention are intended to initiate a public discussion on copyright and AI.

Are there currently already licence negotiations with providers of AI platforms?

GEMA is in dialogue with the relevant AI models and systems providers. It was the first collecting society worldwide to develop a licence model for this purpose. You can find a detailed description of the licence model on our website. However, the licence is not yet established on the market. Many providers want to evade their responsibility by pointing to unresolved legal issues in connection with generative AI, the aim being to operate their business models for as long as possible without paying royalties. They have obtained the raw materials required to operate their business models, namely the works created by humans, from the internet without asking the rights holders for permission and without paying remuneration.

What is the legal situation in the USA?

The legal situation in the USA differs from that in Europe. While it is expressly and clearly regulated in Europe that AI providers may not train AI using protected works for which a reservation of rights has been declared (opt-out), this is highly controversial in the USA. The so-called “fair use” principle applies there, i.e. uses that do not significantly affect the rights holders are permitted without authorisation and without remuneration. Whether this applies to AI training is currently being clarified in a number of US court cases. However, this potential US loophole also appears to be motivating primarily US-based providers to refuse to pay the licence fee owed in Europe, despite the divergent legal situation.

Surely GEMA also uses AI tools? Isn't that a contradiction?

Like many other companies, GEMA uses AI, especially for internal purposes. From GEMA's point of view, it would be negligent not to incorporate the new technology’s possibilities into its work processes. However, unlike AI systems providers, GEMA pays a licence fee to providers for use. GEMA does not want to prohibit the use of its works for artificial intelligence purposes, but rather to licence them so human creators receive fair remuneration. There is thus no contradiction.

Background talk

Any questions?

On November 19 from 10:00 - 11:00 a.m., Dr. Tobias Holzmüller, CEO of GEMA, will be available for a digital background discussion alongside other experts. Please register for the event using the link below. The access link will be sent to you shortly before the event.

Press material

Please direct interview and press inquiries to kommunikation@gema.de

Download

Press photos of the quote givers

Under this link you will find the press photos of Dr. Tobias Holzmüller, Dr. Ralf Weigand and Dr. Kai Welp.

Download press photos

Protect creativity

AI charter

GEMA introduced its AI charter today. It defines ten ethical and legal principles for a fair and sustainable interaction between human creativity and generative artificial intelligence. The AI charter is intended to be thought-provoking and a set of guidelines for a responsible approach to generative AI that respects and protects the rights of creators.

View the 10 principles

article

Licensing model for generative artificial intelligence (AI)

In September 2024, GEMA was the first collecting society worldwide to present a licensing model for generative artificial intelligence (AI). The aim is the fair participation of music creators when their works are used in the training of systems, in the generation of new AI songs or as part of AI-generated music content.

Read the article

Study

Analysis of the impact of AI on the music industry

In January 2024, GEMA published the study “AI and music” with the French CMO SACEM and the research institute Goldmedia. The study focused on the impact of generative AI on the music industry in Germany.

Click here for the study

Dossier

Artificial intelligence in music: information, articles, backgrounds

For all those who want to find out more about AI and music, we offer texts, graphics, background information and press releases on this page.

(Only in German)

Go to the AI dossier

Nimmt die GEMA die Rechte an Songtexten (Lyrics) wahr?

Ja, seit einigen Jahren haben die Mitglieder der GEMA die Möglichkeit, ihr auch die so genannten grafischen Rechte, also die Rechte am geschriebenen Textbild einzuräumen. Die GEMA lizenziert z.B. über ihre Tochtergesellschaften große Musikdienste, wenn diese beim Streamen der Musik auch die Lyrics einblenden.

Now music comes to your mailbox

AI section in the GEMA Newsletter

With the GEMA newsletter, you receive the latest developments in AI and music directly to your e-mail inbox once a month. Also in the newsletter: exclusive interviews, interesting background information, important tips on funding opportunities or current events and more.

Subscribe now

Die GEMA klagt für eine faire Entlohnung.

Ist es richtig, dass die GEMA gegen Anbieter von KI-Systemen Klage eingereicht hat?

Ja, das ist richtig. Die GEMA hat Klage gegen OpenAI Ireland ltd., die Betreiberin des KI-Tools „ChatGPT“ in Europa und deren Muttergesellschaft OpenAI L.L.C. vor dem Landgericht München erhoben. Die GEMA wird prüfen, ob sie künftig noch gegen weitere KI-Anbieter Klagen einleiten wird. Ziel dieser Klagen ist es, eine Lizenzvergütung für die Urheberinnen und Urheber sowie die Musikverlage, auf deren Werken die KI-Systeme beruhen, zu erwirken. Der Erfolg von KI-Systemen, beruht auf den Inhalten, mit denen sie trainiert wurden. Dann ist es nur fair, wenn diejenigen, die diese Inhalte geschaffen haben an den durch KI-Systeme erzielten Erlösen beteiligt werden.

Warum geht die GEMA gerichtlich gegen Open AI vor?

Durch die technologischen Entwicklungen im KI-Sektor gibt es völlig neue Möglichkeiten, Musik zu generieren. Die GEMA befürwortet dies generell. Sie sieht in KI-Systemen ein wichtiges Hilfsmittel im kreativen Prozess. Soweit jedoch Inhalte von ihren Mitgliedern zum Training der Systeme genutzt werden, müssen diese dafür fair entlohnt werden. Die GEMA konnte feststellen, dass bei „ChatGPT“ Songtexte von Werken (Lyrics), deren Rechte sie wahrnimmt, der breiten Öffentlichkeit zugänglich gemacht werden. Das ist nur möglich, weil dieses Tool zuvor mit diesen Texten trainiert wurde. Weder für das Training der KI noch für die Wiedergabe der Songtexte wurde jemals eine Vergütung seitens OpenAI gezahlt. Die öffentliche Zugänglichmachung von geschützten Werken ist aber bereits nach geltendem Recht lizenzpflichtig. Diesen Standpunkt möchte die GEMA zugunsten der Kreativen gegenüber Open AI durchsetzen und hat daher Klage erhoben. Damit ist nicht bezweckt, die Nutzung von Werken der GEMA durch KI-Systeme zu unterbinden. Mit der Klage wird eine faire Beteiligung der Urheber und Urheberinnen an der Nutzung ihrer Werke durch entsprechende KI-Tools angestrebt.