Although AI-generated music has now become commercially mainstream, the legal framework is still struggling to catch up and has not yet kept pace with development. Currently, this industry is not guided by clear rules, but is shaped jointly by litigation, private licensing deals, and regulatory measures in their early stages.

For right holders, platforms, and technology developers, the core contradiction is clear: AI music has already been used and commercialized on a large scale, but its legal foundation remains in an uncertain state. In this article, we will explore the key legal developments surrounding AI music and what exactly these developments mean for businesses operating within or around the music industry.

Litigation: Key Issues Remain Unresolved

Most disputes regarding AI music can ultimately be boiled down to three fundamental issues:

• First, whether training AI models on sound recordings or musical works constitutes an act of "reproduction" under the provisions of copyright law;

• Second, who (if anyone) is the "author" of AI-generated music;

• Third, if AI-generated content is similar to existing works, whether infringement liability will arise.

In the United States, major record companies such as Universal Music, Sony Music, and Warner Music have filed lawsuits against several AI platforms (including AI platforms specifically directed at the music field, such as Suno and Udio, and more recently, Anthropic under which the Claude model is sued), alleging that these platforms used copyrighted music without authorization during the model training process, and expressing serious concerns regarding the generation of infringing copies. Since these lawsuits were filed, some of these cases have reached settlements (particularly those involving Udio and Warner Music), while other cases (especially those involving Suno, Sony Music, and Anthropic) are still ongoing. As of now, US courts have not yet rendered any final judgments to resolve those core legal issues surrounding the relationship between AI music and copyright. The fact that these cases are brought against AI model developers rather than end users who generate the final output may indicate that liability for infringing outputs is more likely to be borne by developers—but this point remains to be determined by final judicial adjudication.

Litigation activity in the UK in this field lags behind the US. Currently, there are still no judgments specifically directed at AI music in the UK, but the High Court's ruling in Getty Images v. Stability AI provides an important contextual reference. The court held in that case that the Stable Diffusion AI model developed by Stability AI did not itself constitute an "infringing copy" because the model did not store copies of the relevant copyrighted works. However, crucially, the judgment did not resolve a broader question of whether using copyrighted works for AI model training constitutes copyright infringement under UK law.

Another more complex issue is the attribution of jurisdiction. If the model training acts occur outside the UK, right holders may face significant evidentiary hurdles and territorial jurisdictional obstacles when initiating successful litigation. This practical situation highlights a practical contradiction: AI development itself possesses a naturally global character, whereas copyright law is established upon strict principles of territoriality. Businesses on both sides of the debate should maintain a sufficiently clear understanding of this gap.

As far as the UK legislative framework is concerned, statutory law does provide certain guidance for these issues. The Copyright, Designs and Patents Act 1988 explicitly recognizes that "computer-generated" works can obtain copyright protection, and defines the author of such works as "the person by whom the arrangements necessary for the creation of the work are undertaken." However, in the absence of case law to interpret how this provision applies to modern generative AI systems, it remains uncertain whether the user inputting prompts or the developer of the AI model qualifies as the "author" under this definition.

The Shift Toward Licensing Models

Major right holders and tech companies are increasingly exploring licensing arrangements rather than relying solely on litigation channels. These licensing arrangements allow music catalogs to be used for AI model training under terms agreed upon by both parties. Such deals typically encompass several elements, such as legally authorized datasets, mechanisms to ensure artist consent or control (sometimes adopting an opt-in model), and some form of compensation or revenue-sharing arrangements, although the specific structure of these elements is still continuously evolving and a unified standard has not yet formed across the industry.

This marks an important shift in thinking: from viewing AI training itself as an inherently infringing act to viewing it as a use that can be authorized through licensing, much like sampling or streaming. In these scenarios, use is permitted, but payment is required for it. For businesses developing music AI tools, obtaining appropriate licenses is rapidly becoming an effective path to mitigate legal risks and safeguard brand reputation.

In the UK, the government's recent decision not to proceed with a proposal that would have substantially expanded text and data mining exceptions (which would have allowed broader AI training provided right holders had the right to opt out) further reinforced the direction of the shift toward licensing models. Instead, recent policy discussions have emphasized the important role of licensing and commercial negotiations as an avenue to facilitate access to protected works. Unlike the broader fair use exception under US copyright law, the UK fair dealing framework is much narrower, which makes licensing a particularly crucial way to use copyrighted works.

However, this model remains highly complex. Key unresolved issues in such licensing deals include: how to value musical works used for model training; whether licenses extend to the authorization of singing styles or "voice-cloning" reproductions; and how AI-generated outputs are utilized outside controlled environments. These are not merely technical issues; they touch upon the core of how revenue is distributed across the AI music value chain and how risk is allocated between various links.

Regulation Remains Fragmented

Although governments are responding, a single, comprehensive regulatory framework has not yet emerged to date. The result is that regulatory measures in different jurisdictions are at different stages of development, presenting a patchwork landscape.

In the European Union, the EU AI Act introduces transparency requirements for certain AI-generated or manipulated content, such as the disclosure of deepfake content and labeling obligations under specific contexts. Enforcement actions are expected to commence later in 2026 once the remaining rules under the Act are implemented. In the United States, there is currently still no comprehensive federal AI law. Instead, regulatory measures are gradually emerging at the state level, with legislators focusing their attention on narrow but rapidly growing risk areas, such as voice cloning, deepfake identity impersonation, and the unauthorized use of another person's likeness or voice.

In the UK, the government has been reviewing how existing legal frameworks (particularly copyright law, data protection rules, and performers' rights) apply to AI-generated content. A key policy focus is the concept of "digital replicas," including AI-generated voice or likeness simulations. The government has conducted extensive calls for evidence and consultations with the creative industries regarding whether new, specialized protective measures need to be formulated. However, a fully implemented digital replica framework does not yet exist, and future rules are still in an active process of formulation.

Alongside government regulation, digital platforms and industry organizations are also introducing their respective governance mechanisms. For instance, Spotify has updated its platform policies to address the issue of AI-generated music, with specific measures including increasing metadata transparency, combating the uploading of spam-like AI content, and participating in industry-level efforts (such as standards formulated by organizations like Digital Data Exchange) to improve the labeling and attribution management of music content.

Overall, the result is a multi-layered, continuously evolving governance system. Within this system, legal regulation, platform policies, and industry standards are developing in parallel, but inconsistencies still exist across different regions and different sectors.

What This Means for Your Business

AI-generated music is neither stopped by litigation nor fully regulated by statutory law. Instead, it is being shaped and defined through a combination of legal pressure, commercial negotiations, and platform governance.

For businesses operating in this field (whether they are right holders, technology developers, platforms, or investors), the key challenge is no longer whether AI-generated music is legal in principle. The real challenge lies in how to structure rights, manage risks, and achieve value capture within a system where rules have not yet taken shape and are still being continuously refined and established. Proactively engaging with and adapting to the emerging licensing and regulatory landscape is crucial.