Your Voice Can Be Cloned in Minutes With AI — So What Rights Do You Actually Have Over It?

AI voice cloning has quietly become one of the most unsettling capabilities of modern artificial intelligence. Tools built on platforms like ElevenLabs and Suno can now recreate a convincing replica of someone’s voice from just a few minutes of audio, no expensive studio equipment or specialized skills required. That’s a serious problem for anyone whose voice exists online in any form — which, in 2026, is most of us.

The issue recently made headlines when Taylor Swift filed trademark applications for phrases closely tied to her voice and public identity, including her signature spoken intro lines. Around the same time, actor Matthew McConaughey pursued similar trademark protection for some of his most recognizable catchphrases. Both moves highlight a question that’s becoming harder to ignore: in an age where anyone’s voice can be synthetically recreated in minutes, what legal protections actually exist — and do they extend beyond celebrities to ordinary people?

To find out, legal experts specializing in technology, intellectual property, and AI law weighed in on where the law currently stands, where it falls short, and where it might be headed next.

Do You Actually “Own” Your Voice?

It might seem intuitive that your voice belongs to you in some legal sense, similar to how you own a piece of writing or a photograph. But according to legal experts, that intuition doesn’t match reality.

Dr. Edina Harbinja, an Associate Professor in Law at the University of Birmingham who specializes in digital rights and emerging technology regulation, explains that voice ownership isn’t as straightforward as it might seem. A specific recording of a voice can be protected through copyright and related rights, and a performance might carry its own performers’ rights — but the voice itself, independent of any particular recording, generally isn’t owned in the same legally concrete way a song or written work is.

Professor Andres Guadamuz, an intellectual property and technology law expert at the University of Sussex, put it even more bluntly regarding the situation in the UK: as things stand, you simply don’t own your own voice. He notes there are workarounds people can attempt, but ultimately there’s little legal recourse available if someone uses your voice to train an AI system, especially when that person had legitimate access to the original recording in the first place.

Why Ownership Might Be the Wrong Question

Rather than focusing on ownership, Harbinja suggests the more useful legal question is whether an AI-generated voice that closely resembles a real person ends up misappropriating their identity, falsely suggesting their endorsement, or misleading the public in some way. In other words, the issue isn’t necessarily who owns a voice — it’s whether a cloned version of that voice causes real-world harm, deception, or misrepresentation.

That framing shifts depending on where you live, since the legal tools available vary significantly by country.

How the US Approaches the Issue

In the United States, this kind of dispute is often handled through the right of publicity, false endorsement claims, and impersonation-style legal theories. A well-known example is the case of Midler v. Ford Motor Co., in which singer Bette Midler successfully challenged Ford’s use of a soundalike performer in one of its advertisements — establishing that imitating someone’s distinctive voice for commercial purposes can create legal liability, even without directly using their actual recorded voice.

More recently, proposed legislation such as the NO FAKES Act has aimed to more directly protect a person’s voice and likeness from unauthorized digital replication, reflecting a growing recognition among US lawmakers that existing publicity-rights frameworks may need updating for the AI era.

How the UK Approaches the Issue

The UK takes a notably different, and arguably weaker, approach. According to Harbinja, there is no general right in the UK to own one’s voice, image, or overall persona. Instead, people seeking legal protection have to rely on a patchwork of related legal concepts — including passing off, performers’ rights, misuse of private information, data protection law, defamation, contract law, and in some cases, trademark law.

Passing off, which protects individuals and businesses from misrepresentation that could damage their reputation or imply false commercial endorsement, is one of the more commonly cited routes. Guadamuz notes that people with an established public reputation can potentially use passing off to stop others from profiting off their image, but its application has historically leaned more toward visual likeness than voice, and its scope remains limited.

Harbinja summarizes the core problem well: voice cloning simply doesn’t fit cleanly into any single existing legal category. Depending on the country and context, an instance of AI voice imitation might be argued as impersonation, false endorsement, personality infringement, unlawful data processing, unfair competition, a performers’-rights violation, or straightforward consumer deception — with no consistent, unified legal framework tying these approaches together.

The Celebrity vs. Everyday Person Problem

This is where Taylor Swift’s trademark strategy becomes particularly interesting. Trademarking a specific phrase associated with her voice can offer some protection in commercial contexts — for example, preventing a company from using an AI-generated version of her saying “Hey, it’s Taylor” in an advertisement without permission. But trademark protection is narrow by design. It doesn’t solve the broader, more troubling issue of unauthorized AI voice imitation outside of commercial branding situations.

And if a celebrity with significant legal resources and a globally recognized personal brand still has to rely on narrowly scoped legal tools like trademark filings, the situation is considerably murkier for ordinary people. Most individuals don’t have a trademarked voice, an established public reputation to invoke through passing off, or the financial resources to pursue a legal claim even if one were theoretically available.

Two Structural Problems With AI and Voice Cloning

Harbinja identifies two separate enforcement challenges that make this issue particularly difficult to address through existing law.

The first is a training-data problem. AI voice models are frequently built using enormous collections of recordings scraped from across the internet — interviews, performances, podcasts, videos, and more. Rights holders often have no way of knowing whether their specific recordings were used to train a given model, where that training took place, or whether a legally significant copy was made in a jurisdiction where they’d even have standing to sue.

The second is an output problem. If an AI-generated voice sounds like a specific public figure but doesn’t reproduce a substantial, identifiable portion of any single protected recording or composition, then traditional copyright law may simply not apply. That gap is part of why the broader legal conversation is shifting away from copyright as the primary tool, and toward newer concepts centered on voice, likeness, personality rights, publicity rights, informed consent, transparency requirements, and regulation of digital replicas more generally.

Guadamuz echoes this concern from an enforcement standpoint, noting that while the UK offers essentially no protection and other countries offer somewhat more, actually enforcing any of it in practice remains extremely difficult. In his view, only the most blatant, egregious cases of voice cloning are likely to result in any real legal action.

Is New Legislation on the Way?

Slowly, yes. Regulatory frameworks are starting to catch up, even if they’re not yet comprehensive.

Harbinja points to the EU AI Act as one meaningful step. While it doesn’t create an ownership right over a person’s voice, Article 50 introduces transparency requirements for certain AI interactions, including deepfakes and synthetic audio content — meaning AI-generated voice content may increasingly need to be clearly disclosed as artificial.

Guadamuz believes that piecemeal legal workarounds ultimately won’t be enough, and that dedicated legislation specifically protecting voice will eventually be necessary. He notes that UK policymakers have already begun discussing broader protections against deepfakes and digital replicas — discussions that could plausibly expand to directly cover AI-generated voice cloning in the future.

A Deeper Question Than Ownership

Harbinja raises a compelling philosophical point that goes beyond the technical legal mechanics: should voice, likeness, and personal identity be treated primarily as property to be owned, or should they instead be governed through concepts like dignity, personhood, and individual autonomy? Her own view leans toward the latter, particularly in cases where AI systems aren’t merely copying a commercially marketable asset, but actively simulating a real, identifiable person.

That distinction matters because it reframes the entire debate. This isn’t purely a dispute about copyright, trademarks, or data protection compliance — at its core, it’s a question about identity and authenticity in an age when a voice, one of the most fundamental ways people recognize each other, can be convincingly recreated by a machine in a matter of minutes.

What This Means for You

If you’re not a celebrity with trademark lawyers on retainer, the honest answer is that your legal protections against AI voice cloning are currently limited and inconsistent, especially outside the US. Depending on where you live, you might have narrow recourse through data protection law, defamation, or misuse-of-private-information claims if a cloned voice is used to harm or deceive you — but there’s no broad, dedicated legal shield covering voice cloning the way copyright covers a song or a piece of writing.

For now, the most realistic protections come down to platform-level safeguards, emerging transparency requirements like those in the EU AI Act, and the slow, ongoing process of lawmakers catching up to a technology that’s advancing far faster than legislation typically moves.

Frequently Asked Questions

1. Can I legally stop someone from cloning my voice with AI?

It depends heavily on where you live and the specific circumstances. In the US, tools like right-of-publicity and false-endorsement claims may offer some protection. In the UK, there’s no general right to your own voice, and available protections are limited to a patchwork of related legal concepts like passing off and data protection law.

2. Why did Taylor Swift trademark phrases like “Hey, it’s Taylor?

Trademarking specific spoken phrases gives her a narrow but concrete legal tool to stop commercial misuse of AI-generated versions of her voice using those exact phrases, even though it doesn’t protect her voice more broadly.

3. Does copyright law protect my voice?

Not directly. Copyright typically protects a specific recording or song, not the underlying sound of a person’s voice itself. If an AI-generated voice sounds similar to someone but doesn’t reproduce a substantial part of any specific protected recording, copyright law usually doesn’t apply.

4. What is the NO FAKES Act?

It’s proposed US legislation designed to more directly protect a person’s voice and likeness from unauthorized AI-generated digital replicas, reflecting growing concern among lawmakers about the gaps in existing publicity-rights law.

5. What is passing off, and how does it relate to voice cloning?

Passing off is a UK legal concept that protects people and businesses from misrepresentation that could damage their reputation or falsely imply endorsement. It’s one of the few legal routes available for voice-related disputes in the UK, though it was historically designed more for visual likeness and has limited application to voice cases.

6. Does the EU AI Act protect voice cloning victims?

It doesn’t create an ownership right over your voice, but Article 50 introduces transparency requirements for certain AI-generated content, including synthetic audio and deepfakes, meaning such content may need to be clearly disclosed as AI-generated.

7. Are ordinary people (not celebrities) protected from voice cloning?

Generally, protections for everyday people are even weaker than for celebrities, who at least have options like trademark filings and public reputation to draw on through passing-off claims. Most individuals have limited practical recourse under current law.

8. Is new legislation likely to change this in the future?

Legal experts believe dedicated legislation specifically addressing voice cloning will eventually be necessary, and note that UK policymakers have already begun discussing broader protections against deepfakes and digital replicas that could extend to voice cloning.