In recent decades, artificial intelligence, often abbreviated as “AI,” has had a huge impact on various industries, including the music industry. With the rise of advanced algorithms and machine learning technologies, musicians and music platforms have begun to embrace AI. This adoption aims to improve creative processes, make music creation more efficient, and enrich the experience of music lovers. But, in addition to all the positive developments, these advances also raise a series of legal and philosophical issues. Of particular interest for this article are the legal implications of AI in the music industry.

Intellectual Property and AI in the Music Industry

One of the main legal issues arising from the use of AI in the music industry pertains to intellectual property. AI has changed the way we compose, produce, and distribute music. While AI opens the door to new creative possibilities, it also raises important questions about protecting the rights of musicians, composers, and producers. In this article, we will explore the essential concepts of music copyright and related rights. Then, we will briefly discuss the impact of AI on these rights.

Copyright and Artificial Intelligence

Copyright is a legal concept granting exclusive rights to the creator of an original work, including musical compositions and lyrics. It gives the author the right to determine how the work is used, reproduced, distributed, and presented to the public. With the advent of AI in the music industry, the question arises whether this algorithmically created music also qualifies for copyright protection.

In addition, we must consider whether the output has infringed on the material used as training data by the AI model. And, if so, whether there may be exceptions to the rule.

Another challenge with AI music is answering the question of whether it is a “work” within the meaning of the Copyright Act and then determining its creatorship. Traditionally, the human composer is the author of a piece of music. But, with AI having the ability to create unique compositions, the question becomes relevant: who is the real creator?

I asked ChatGPT a question about this. Unsurprisingly, it replied: Legal systems may require adjustments to recognise AI as ‘authors’, or to consider those who created the AI as authors. So, AI is already claiming rights!

This question is also relevant for publishers and collective rights organisations. Because even if you can trade in compositions and texts, if there are no rights attached to them because the creator is not human, then it is also challenging to make money from them. That money is earned due to the exclusivity of control over the works and the licensing fees paid to use them. No right means no exclusivity. No exclusivity means no, or at least much less, money.

The Question of Authorship and Originality

The European Union faces challenges in adapting its legal framework to the unique nature of AI creations. In this context, the EU has taken steps by enacting the EU AI Regulation. This aims to establish harmonized rules for AI development and usage across EU states, focusing on safety and fundamental rights. However, it’s notable that the regulation doesn’t directly address copyright and generative AI in the music industry. Instead, it prioritizes transparency and accountability in AI systems, setting a precedent for future legislative developments. It provides a broader AI governance framework, suggesting that specific copyright issues may need to be addressed later.

The question of creatorship in AI-generated music becomes increasingly complex with advanced AI systems. This becomes particularly relevant in the context of copyright laws, which traditionally recognize human creators.

Determining originality in AI-generated music involves assessing whether the work reflects the author’s free and creative choices, particularly under EU law. This can be challenging due to AI’s iterative nature, like machine learning, which often obscures direct human contribution. The “author’s own intellectual creation” standard requires a higher originality threshold, often beyond AI’s capability without human intervention. Legal presumptions about authorship and ownership further complicate this aspect. Such assumptions may not always accurately determine the originality of AI outputs​​.

Works like AI-generated songs by AIVA, where user or programmer input significantly shapes the outcome, blur the lines of authorship. Under English law, as seen in s. 9(3) of the Copyright, Designs, and Patents Act 1988 (hereafter, “CDPA”), the author is considered to be the person responsible for the creation of the work. This raises complex questions when AI tools are involved, particularly in scenarios where users input prompts or commands. The degree of human intervention in AI-generated music directly impacts the assessment of originality and authorship​​.

Examples of Copyright in AI Art and Music

A notable example that sheds light on this issue is found in the visual arts field. An AI program learned to paint in the style of the famous Dutch artist Rembrandt. The AI did not act independently; rather, it was heavily guided by human intervention. Artists and technologists meticulously programmed the AI to analyze Rembrandt’s works and replicate his style. This case, detailed in this article, illustrates how the application of AI in the music industry, when used as a tool under close human supervision, could potentially yield copyright-protected works. The AI here is not the sole ‘author’; instead, it acts under the guidance and creative direction of humans. This suggests that the final work could be eligible for copyright protection based on the depth of human involvement.

AI tools like AIVA, designed as “Creative Assistants for Creative People,” exemplify the diverse capabilities of AI in music creation. These tools can produce initial drafts, which often require human input for refinement. For instance, Huawei’s AI, which ‘completed’ Schubert’s Unfinished Symphony, worked in tandem with composer Lucas Cantor to enhance its output. This collaborative approach raises questions about the extent of copyright protection for AI-assisted creations. By analogy, in cases like Hyperion Records v Sawkins, the human augmentation of AI-generated content was deemed original. This indicates that similar protections could potentially apply to AI-assisted music compositions​​.

Related Rights for AI in the Music Industry

While copyright protects composers, neighbouring rights protect those involved in making a recording, such as the artists and the producers. You can read exactly how master rights work here .

As AI in the music industry becomes more prevalent in music production, questions may arise regarding the granting of neighbouring rights. If an AI system creates the music, who is the holder of neighbouring rights? Is it the musicians who trained the algorithms or the owner of the AI platform? Or, perhaps, the writer of the AI code?

We can break this query into several considerations:

The Musicians Who Trained the Algorithms

If musicians actively contribute to training the AI—by inputting style, genre, or other musical elements—they could potentially claim some neighbouring rights. Their creative input and expertise guide the AI’s output, making their contribution significant.

The Owner of the AI Platform

The platform owner might argue for neighbouring rights based on their ownership and maintenance of the AI technology. However, this claim is more tenuous as the platform owner’s role is often more administrative and less creatively involved.

The Writer of the AI Code

Coders and developers who create the AI’s algorithms play a critical role in the music generation process. Their argument for neighbouring rights may be that their code is the foundational element enabling music creation.

Copyright Holders of the Source Works

In cases where AI-generated music relies on existing works (either directly or stylistically), the original copyright holders may have a claim to neighbouring rights. This is particularly relevant in jurisdictions like the US. There, the imitation or ‘sound-alike’ versions of a work can infringe on the original copyright holder’s rights. This claim originates from the idea that AI, by using existing works as training data, could produce outputs that are substantially similar to these works in style, sound, or artistic expression. The utilisation of these original works as source material or stylistic inspiration in the AI’s output may grant these holders a stake in the AI-generated content’s neighbouring rights. This is especially true if the AI’s output is substantially similar to or derivative of the protected work.

For the master owner, there is no creative threshold after which a right arises, such as with copyright. It is ‘only’ about the initiative and responsibility for a recording. This raises interesting ownership questions. Because: Who should you ask for permission to sample a track or use the music in a video? If we arrange this properly with the metadata, we could solve the spaghetti of rights and create a one-stop shop for arranging licenses for music use.

Infringement of Inputs: Text and Data Mining (TDM)

One way in which AI in the music industry can learn to imitate musicians’ voices or compositional styles is by training on large amounts of data, known as “text and data mining” (TDM). The concept of copyrighted AI music becomes intricate when considering the use of TDM for training AI systems. TDM can lead to copyright infringement. Especially when making permanent copies of recordings without appropriate licenses in jurisdictions without a broad statutory TDM exception. Conversely, the position is less clear where only making temporary and transient copies of songs. And, also, where only abstracted parameters are stored and used by the AI model, which are not themselves copyrighted works.

The legal landscape varies globally, with countries like the UK and Singapore having differing approaches to TDM. UK law currently permits “text and data analysis” only for non-commercial research (s. 29A, CDPA). However, in June 2022, the UK Intellectual Property Office announced a proposal to allow TDM for any purpose at all. This exception would allow the training of commercial AI tools on all music without requiring a license or compensating rights holders. This would have made the UK one of the most permissive places for AI research in the world. This received significant objection from the music industry, which described it as “music laundering.” At the opposite end, Singapore has recently enacted a very broad TDM exception.

Global Perspectives and the Future of TDM Legislation

The proposed changes in UK law highlight the ongoing tension between AI innovation and copyright protection. These changes were eventually scrapped in February 2023 due to industry pushback. By comparison, the EU Digital Single Market Directive permits rights holders to opt their works out of the TDM exception.

In such a complex environment, obtaining licenses for TDM becomes crucial to avoid legal disputes. This is exemplified by arrangements like Hipgnosis’s deal with Reactional Music, indicating a move towards more structured licensing frameworks​​.

In addition to commercial licenses freely negotiated between users and rights holders, state-approved licensing schemes may emerge. On March 15, 2023, the report of Sir Patrick Vallance on the Pro-Innovation Regulation of Technologies Review stated that the UK “should enable mining of available data, text, and images (the input).”

The Government’s response stated that, to provide clarity, the UK Intellectual Property Office will produce a code of practice by summer 2023. It also stated that “an AI firm which commits to the code of practice can expect to be able to have a reasonable license offered by a rights holder in return” (see recommendation 2 of the response). As such, the Government seems to be encouraging an industry-led approach to establishing an official licensing framework, with legislation only to be brought in if this cannot be agreed upon.

Voice Clones

Deepfake technology, a notable byproduct of AI in the music industry, raises additional legal questions. Under English and EU law, the style of singing, even if AI generates it, is unlikely to earn copyright protection. However, there could be potential for ‘passing-off’ claims. Especially if an AI synthesizer imitates the voice of a well-known artist for commercial purposes. This is an evolving area of law. As such, recent cases help expand the understanding of copyright and personality rights in the context of AI-generated content​​​​.

Additionally, Article 52 of the proposed EU AI Regulation addresses the need for transparency when using deepfake technologies. It mandates clear disclosure when AI generates or manipulates image, audio, or video content that materially misrepresents reality. This provision aims to prevent deceptive uses of AI. Particularly in cases where deep-fake voices or images could mislead the public or infringe upon the rights of individuals.

Case Studies

There may be an argument that imitative vocal synthesizers could be used in certain ways, which could constitute passing-off. For example, Californian law recognizes that when the “distinctive voice of a professional singer is widely known and is deliberately imitated in order to sell a product, the sellers have appropriated what is not theirs and have committed a tort” (Midler v Ford Motor Co.).

Interestingly, Rick Astley has recently brought a similar case for the imitation of his voice in an interpolation of his track ‘Never Gonna Give You Up’ by Yung Gravy (Astley v Harui PKA Yung Gravy). In the pleadings, Astley’s legal team has sought to expand the Midler judgment to apply to the use of imitation for any commercial purposes, rather than solely in relation to false endorsement.

Whilst Yung Gravy used an Astley impersonator rather than an AI tool, Astley’s case may set a precedent. If successful, it could open an avenue under Californian law for actions against vocal imitations made by AI.

To prevent false celebrity endorsements, the UK used the action for passing off (Irvine v TalkSport). And it may be sufficiently flexible to take action against deep-fake versions of artists’ voices in many cases.

The existence and scope of personality rights varies significantly between jurisdictions. So, attempts to assert personality rights internationally may have varying degrees of success.

Conclusion

To tackle the complex issues surrounding music copyright and neighbouring rights, as influenced by AI in the music industry, collaboration between legal experts, technological pioneers, musicians, and music companies is necessary. A multidisciplinary approach will help shape a balanced and future-proof legal structure. If lawmakers, lawyers, and judges do not understand how technology and music work, the system remains reactive and slow. Conversely, if technicians and musicians are unfamiliar with the law, it hinders their ability to effectively connect with ongoing changes.

Virtually all legal and intellectual property courses now include an element of artificial intelligence. Hopefully, this article will ensure readers of music websites are also aware of what is happening in the legal field. For guidance on specific legal inquiries or issues pertaining to the use of AI in the music industry, don’t hesitate to reach out to Backstage Legal. Together, we create the future, and that sounds like music to our ears.