Austria: The rise of AI-generated music – a challenging legal framework

At first glance, the use of AI music promises a cost-effective and low-risk alternative to traditional licensing. However, the copyright issues are far more complex than assumed. AKM and Austromechana have warned of uncritical use and pointed to their comprehensive licensing competence.[2] Understanding the underlying issues and how to use AI music with minimal risk is essential.

Two types of AI music and their copyright assessment

From a copyright perspective, a distinction must be made between two types of AI music. The criteria for this distinction are the extent and nature of the user's involvement in the creative process.

AI music in the narrower sense is music in which the human input (prompt) remains so generic that it shifts into the background with regard to the piece of music produced.

• Example: Generate a piece of classical music suitable for playing as background during dinner in a fine restaurant.

In this case, the AI acts as the sole "creator".  Consequently, the output is not protected by copyright, as it lacks the mandatory requirement of human authorship. Pure AI music is therefore in the public domain, with any usage restrictions arising, if at all, from contracts with providers. What could be discussed in these situations, if anything, is a property right to the generated sound file (sound recording). However, this ancillary-rights perspective on AI music is currently still scarcely discussed in literature and practice.[3] 

This type of AI music must be distinguished from so-called AI-assisted music. Here, AI is used merely as inspiration or as a tool, while human input shapes the creative characteristics of the resulting piece. A wide variety of scenarios exists in the field of AI-assisted music. For example, all common AI music generators allow users to upload their own lyrics, provide specific prompts, or extensively edit the output. Human creativity can therefore be incorporated at every stage and the resulting work may then be protected by copyright.

However, the boundaries are fluid and difficult to define with legal certainty.

When using existing pieces of music and sound sequences labelled as "AI music", for example on licensing platforms, it is rarely possible to trace the creation process. However, in general the responsibility for any potential infringement lies with the user. It is therefore important to carefully check the licensing and terms of use of the platforms.

Collecting societies and their role in the exploitation of AI music

Collecting societies such as AKM generally act as administrators of works in the music sector that are protected by copyright and neighbouring rights. They issue the licences required for commercial use at industry-specific individual or collective tariffs. Traditionally, it is assumed that collecting societies manage and license a "world repertoire" through agreements with affiliated societies.

Accordingly, there is a (rebuttable) legal presumption in Austria that any publicly accessible music belongs to the AKM repertoire. Therefore, it is the user's responsibility to prove that the music is "AKM-free".

In addition, works by rights holders who have not concluded a representation agreement can also be effectively licensed for use in the catering and retail sectors through a so-called extended collective licence, as long as the right holders do not expressly object. This is intended to promote legal certainty and simplify the licensing market.

What does this mean for AI music?

Pure AI music is not created by a human who could be a member of a collecting society, and according to prevailing opinion, it is not protected by copyright. Consequently, pure AI music has nothing to do with collecting societies. However, due to the AKM presumption, anyone who publicly reproduces AI music must still prove, in the event of a dispute, that no member of a collecting society was creatively involved.

For AI-assisted music, however, the rules are the same as for any other music: if the human creators are members of a collecting society, the AI-assisted music falls within their repertoire; if they are not members, the collecting societies can only issue extended collective licences.

What's next?

It seems questionable whether the AKM presumption can still be justified in view of the increasing relevance of AI music. At present, it is leading to considerable legal uncertainty. It would be desirable to develop a procedure for distinguishing between music that is subject to licensing and music that is not, and to define the simplest possible form of evidence that collecting societies will accept in this regard.

Additional legal issues regarding the use of AI music

There are further legal questions surrounding AI music that may become critical depending on the use case and perspective. These include, for example, the use of existing works for AI training, the use of existing works within user prompts (input), or liability issues if the AI-generated piece of music recognisably contains elements of existing works (output).[4] While the first two problems mainly affect AI operators and distributors of AI music, the output side also poses risks for mere users of "purchased" AI music.

In addition to copyright, personal rights may also be infringed if, for example, the voices of real people are used without their consent. When used in advertising, the AI Act, which will come into force in 2026, may also require labelling as an "AI output" in specific cases.[5]

authors: Dominik Hofmarcher, Marie-Therese Wirtz