Artificial intelligence and music might not sound like a match made in harmonic heaven, but the fact is that AI is reshaping how songs are created, promoted and discovered. Models can now compose melodies, generate convincing vocals and imitate the stylistic signatures of well known artists with striking accuracy. This shift brings opportunity but also exposes deep tensions in the legal structures that protect artists and the commercial ecosystem around them.
The rise of the AI generated band Velvet Sundown and the dispute involving producer Haven and Jorja Smith show how quickly the boundary between inspiration and imitation is eroding. As AI becomes embedded in the creative process, the industry is being forced to confront urgent questions about authorship and accountability.
A legal framework under strain
In the UK, copyright protects musical compositions, lyrics, sound recordings and printed editions from the moment they are created. Artists also build protectable goodwill in their name and reputation, as seen in the Robyn Rihanna Fenty v Arcadia Group Brands Ltd case, which confirmed that passing off can prevent misleading commercial use of an artist’s identity.
This framework was not designed for generative AI systems trained on vast quantities of copyright-protected material. If a model ingests thousands of tracks and later produces something close to a protected recording, the question becomes who is legally responsible. The developer who trained the model, the user who generated the output or the platform that hosts it.
These questions are no longer theoretical. They are central to how the industry will evolve.
The Jorja Smith and Haven dispute
A recent high profile dispute so far concerns Haven’s track I RUN, which went viral on TikTok before attracting millions of streams on major platforms. Smith’s label Famm alleges that the vocal performance was generated using AI trained on Smith’s music and that promotional posts tagged with her name contributed to public confusion.
Haven denies using Smith’s recordings and says the track was created using their own vocals processed through Suno. Suno has acknowledged that its systems are trained on copyrighted works and has argued that this falls within lawful exceptions. The track was later removed from charts and platforms following takedown notices.
Famm has suggested that both copyright infringement and passing off may apply. The case illustrates how easily AI assisted works can create confusion about an artist’s involvement and how quickly misinformation can spread online.
Training data, infringing outputs and platform liability
The legal uncertainty is heightened by the lack of case law. Key questions now emerging include whether training a model on copyrighted works without permission is itself an infringement and whether responsibility for infringing outputs sits with users or developers.
The Getty Images v Stability AI judgment provides some insight. Although many issues fell away due to jurisdiction, the court found that it was the training of the model that caused Getty’s watermark to appear in outputs. This suggests that platform developers could be held responsible when trained systems generate infringing material.
If this reasoning were applied to music, it could shift enforcement efforts toward AI developers rather than the individuals using their tools. And that fundamentally changes everything, blaming the provider rather than the end user.
Suno, and other platforms could turn toward licensing
Litigation has already begun in the industry. Suno faced lawsuits in the United States from Sony Music, Warner Music Group and Universal Music Group for allegedly scraping copyrighted tracks to train its system. Damages sought reached up to 150,000 dollars per infringed work – a hefty exposure in damages when considering the sheer number of works that would be used to train a model.
The landscape changed when Warner Music Group reached a licensing agreement with Suno, allowing its artists to opt in to training and providing oversight of outputs. What this move favours is controlled participation and agreement, over prolonged litigation, signalling a potential solution could be commercial licensing in model training.
Government intervention remains unsettled
The UK government has considered a text and data mining exception that would allow developers to train models on copyright-protected content when they have lawful access, subject to an opt out. The music industry has strongly opposed this approach, citing the imbalance it creates.
A House of Lords inquiry is now underway to examine how rightsholders can effectively reserve and enforce their rights. Evidence presented by UK Music’s chief executive has emphasised the cultural and economic importance of copyright to the sector.
Toward a framework that supports innovation and authorship
Generative AI will continue to grow in capability and influence, in music and beyond. AI produced vocals and compositions are already difficult to distinguish from human performances and will only become more convincing. Without intervention, artists risk losing control over their sound and audiences risk losing clarity about what they are hearing.
Potential solutions include clear labelling of AI generated music, industry wide licensing frameworks for training data, defined liability rules for infringing outputs and shared standards that reflect both creative ethics and commercial realities.
A combination of legislation, case law, commercial agreements and industry norms will be needed. No single mechanism can resolve the challenge alone – it’s too complex for that.
A new creative coexistence
Despite the panic, we don’t believe that AI will replace human creativity and music creation, but it certainly will continue to reshape the environment in which musicians operate. The challenge is to build a system that protects artists while allowing innovation to flourish. That requires clarity, fairness and accountability, the principles that have always underpinned copyright.
The future of music will be hybrid, part human and part algorithmic. The priority now is ensuring that the people whose creativity built the industry remain protected and empowered as that future unfolds. Achieving this will require a robust and future-proof legal framework.
About Edwin Coe
Intellectual property law is a vital tool in the creation and protection of dynamic business assets and Edwin Coe’s Intellectual Property team combines dedicated IP specialists and heavyweight litigators to provide a complete range of IP-related services.
Our substantial team comprises solicitors, barristers and trade mark attorneys and we advise on the acquisition, exploitation and protection of intangible rights.
https://www.edwincoe.com/services/intellectual-property/
