The sound of ownership: why AI voice cloning is raising new IP questions

As AI voice cloning technology becomes increasingly sophisticated and accessible, questions surrounding the protection of an individual's voice are moving to the forefront of legal debate. Against this backdrop, the Save Our Voices Now campaign has attracted support from a number of high-profile celebrities, calling for a statutory right allowing individuals in Britain to own and control their voice as intellectual property.

As AI-generated voices become increasingly indistinguishable from the real thing, the campaign raises a pressing question: does existing law sufficiently protect for one of our most personal and recognisable attributes?

In this article, we examine the current legal framework surrounding voice protection, consider where existing rights may apply, and outline the practical steps individuals and businesses can take to protect themselves.

The rise of AI voice cloning

AI voice cloning uses machine learning to replicate a person's voice from audio samples, generating new speech that can reproduce characteristics such as tone, pitch, accent and speaking style. As this process begins with recordings of an individual’s voice, it raises difficult questions around intellectual property ownership and enforcement.

The campaign seeks to address perceived gaps in protection through three principal objectives:

introducing a statutory right for every person to legally own their voice as their personal intellectual property;
placing the legal burden of proving consent for the use of a voice on any platform on the AI developer, not the individual challenging usage; and
making the right enforceable in UK courts against any platform serving UK users, regardless of where the platform is based.

The current legal framework

The Copyright, Designs and Patents Act 1988 establishes that copyright does not subsist in literary, dramatic and musical works until they are recorded, in writing or otherwise. Crucially, therefore, copyright does not protect a person’s voice as such. Copyright may protect a recording containing that voice, but the relevant intellectual property right sits in the protected work or recording, rather than the personal attribute of the speaker’s voice itself.

This exclusion of the spoken voice from current statutory framework highlights a fundamental issue raised by the campaign: the existing intellectual property framework does not provide a straightforward proprietary right in a person's voice itself.

Objective one: legally protecting people’s voices as their personal IP

The campaign's first objective is rooted in the idea that individuals should control the use and replication of their own voices. However, this raises questions around whether the campaign aims to protect personal rights relating to an individual’s voice or whether they are trying to protect the proprietary right to monetise it.

Other jurisdictions offer broader protection. In the US case Midler v Ford, Bette Midler successfully challenged the intentional imitation of her voice in a Ford advertisement under US personality-based protections, rooted in the torts of right to privacy and right to publicity. The court accepted evidence that viewers were mistakenly convinced that Midler had sung in the advert, when in reality Midler had refused to partake. However, the UK does not recognise any equivalent standalone personality rights.

Despite the absence of a general right protecting an individual’s voice in the UK, existing intellectual property rights may offer some protection where particular elements of a voice have acquired a distinctive character.

Copyright

Copyright may offer some protection where AI voice cloning involves the use of an existing sound recording. Copyright can subsist in the sound recording itself, meaning that use of that protected material to train AI voice cloning technology without permission may amount to infringement. However, this protection attaches to the recording, rather than to the voice captured within it. As a result, the person whose voice is heard may not own the relevant copyright, for example if the recording is from a movie, then intellectual property rights may be owned by the producer. Therefore, the targeted individual may not be entitled to bring a copyright claim themselves.

There is also greater uncertainty where AI merely replicates the characteristics of an individual's voice without reproducing protected elements of the original recording. In those circumstances, copyright may offer limited protection because it does not provide a general proprietary right in the sound or characteristics of a person's voice.

Trade mark protection Registering one's voice as a registered trade mark is another possible route to voice protection. Matthew McConaughey has secured such protection in the US covering aspects of his voice and likeness as part of an attempt to strengthen protection against unauthorised AI use. Specifically, this covers his distinctive delivery of the iconic phrase “aright, alright, alright” from his breakout role in Dazed and Confused (1993).

However, this narrow protection illustrates a potential limitation of trade marks in the context of voice cloning. A sound mark protects the particular mark registered, rather than granting a general monopoly over the individual's voice. This raises a difficult question: if AI replicates the recognisable characteristics of a person's voice, but uses that voice to say entirely different words, will it amount to use of the registered mark?

Passing off

Unlike copyright, passing off does not depend upon ownership of a particular recording. Instead, it may offer protection where an individual has established goodwill in commercialising their voice, a third party makes a misrepresentation that leads the public to believe that the individual has endorsed or is otherwise commercially connected with a product or service, and damage results.

Consider McConaughey again, whose voice has become synonymous with audiobook recordings. If an AI-generated imitation were used to narrate a new work in McConaughey's voice, an argument might arise that McConoughey has goodwill in connection with his narration work and that use of a sufficiently recognisable imitation misrepresents that McConaughey is associated, has endorsed or is involved in the project. Potential damage could include lost commercial opportunities or reputational harm, particularly if, for example, the work in question is something derogatory which McConaughey would not want to be associated with.

Objective two: placing the legal burden of proving consent for the use of a voice on any platform on the AI developer

The campaign's second objective would require platforms facilitating voice cloning to demonstrate that valid consent was obtained. This could address the practical difficulty that individuals may not immediately become aware that their voice has been replicated or know how the relevant AI model was trained. From an enforcement perspective, requiring developers to retain and provide evidence of consent may, in theory, provide a more effective safeguard, however its practical enforceability remains in question.

In relation to their proposed reform, which aims to amend national copyright rules and introduce voice protection, Danish lawmakers have considered the enforcement restrictions on AI voice cloning. They have proposed a notification based approach, with the burden of proof remaining with the individual to raise and establish misuse. Ultimately, this highlights the tension between the campaign's proposed safeguards and the practical realities of enforcing rights against rapidly evolving technologies. Although shifting the legal burden may strengthen an individual’s rights in principle, its effectiveness will ultimately depend on the development of enforcement mechanisms.

Objective three: making the right enforceable in UK courts against any platform serving UK users

The internet is borderless. An individual in one jurisdiction may have their voice cloned by a developer operating thousands of miles away. Although granting UK court’s jurisdiction over platforms serving UK users could strengthen domestic protection, enforcement against overseas developers without meaningful UK assets or presence may remain difficult.

It is unclear whether national legislation can realistically regulate the global AI ecosystem. The Government have acknowledged that leading AI developers do not always train their models in the UK; since copyright is territorial, developers are therefore not required to respect intellectual property rights under UK law. While international agreements such as the Berne Convention help ensure that UK copyright works receive protection in other member states, they do not create a uniform global copyright regime, meaning the legality of AI training activities will still depend to an extent on the laws of the country in which they occur.

How can you navigate AI voice cloning under the current framework?

Whilst the campaign’s objectives continue to gain support, the current framework is not designed to protect the spoken voice as their intellectual property. However, this does not mean that individuals and businesses are without protection. Depending on the circumstances, you may be able to rely on existing intellectual property rights outlined above.

On the other hand, for developers of AI technology it is increasingly important to maintain consistent and comprehensive records relating to the material used to train voice cloning models. Developers should maintain clear evidence regarding the source of training data and the permissions obtained for its use. The fact that audio content is publicly accessible online does not necessarily mean it is free to use; copyright, trade marks and other legal rights may still apply.

Ultimately, the campaign reflects a broader debate about whether the law should evolve to recognise a person's voice as a protectable proprietary asset in its own right. Whether legislative reform is capable of taking the form proposed by the campaign remains uncertain. What is clear, however, is that advances in AI voice cloning technology are testing the boundaries of the existing legal framework. Until greater legal certainty is provided, individuals and businesses alike should understand the intellectual property rights that may already exist and take proactive steps to safeguard their interests.

For more information, please contact our intellectual property team.