Who owns your voice? AI cloning turns our most intimate trait into an asset with an owner

THE THESIS AND THE CASE BEHIND IT. For centuries, a voice could not be separated from the person. That era is over. A few seconds of audio are now enough to build a convincing clone. The question is no longer whether a voice can be copied. It is who gets to decide what happens to that copy. Our thesis is simple: the voice is becoming an asset with an owner, like a trademark or a song. That brings real short-term harm, especially for dubbing, voice-over work and older people. But it also points to a sensible future: every person as the legal owner of their voice, free to license it or refuse. Today's problem is not the technology. It is that consent still has no infrastructure.

In early October, Claudio Serrano posted a video on TikTok. Serrano has voiced Batman in Spain since 2005, as well as Ant-Man and actors such as Christian Bale and Ben Affleck. According to his account, reported by Infobae, a US AI platform displayed a profile with his name, his photo and a synthetic voice sample presented as his. The platform has not been publicly named, and no response from it has been reported. Serrano says his lawyers are working on the case. The most telling detail is another one. According to the actor, to file a complaint he was asked to pass a 'voice CAPTCHA', which meant recording more samples of his own voice to prove it was him. Serrano compared it to asking a victim to be attacked again to prove the attack. Our reading: the anecdote sums up the problem. Verification systems are built to protect the platform, not the person who was cloned. And the victim must take on a company in another country, under another law.

THREE LIVES OF ONE VOICE. Three things get mixed up in this debate, and they should be separated. First, the voice as biometric data. In the European Union, a voice is personal data. When it is used to identify someone, it becomes biometric data, a specially protected category under the General Data Protection Regulation (GDPR). Second, the voice as a scam tool. This is the 'grandchild in trouble' call, where a fake relative uses a cloned voice to ask for urgent money. According to the FBI's 2025 Internet Crime Complaint Center (IC3) report, complaints mentioning AI totaled about 22,000, with around $893 million in losses. It was the first year the FBI tracked AI explicitly, and the figure only counts reported cases. Third, the voice as a product. ElevenLabs, the largest synthetic voice company, launched a marketplace of licensed 'iconic' voices in November 2025. It includes Michael Caine, who is alive, and voices of deceased figures managed by estate agencies such as CMG Worldwide. The company says every commercial use needs the rights holder's approval. Our reading: all three lives run on the same technology. Only one thing separates them: consent. With it, a voice is a legitimate business. Without it, it is fraud or plunder.

THE PROFESSION THAT PAYS FIRST. Dubbing and voice-over are the front line. A 2024 study by PMP Strategy for CISAC, the global confederation of authors' societies, estimates that audiovisual creators have 21% of their revenue at risk by 2028. For translators and adapters in dubbing and subtitling, the figure rises to 56%. It is a projection, not a count, but it shows the direction. In Spain, PASAVE, the platform of voice actors' associations and unions, gave studios a contract clause in 2023 that bans using recordings to train or simulate voices with AI. Its spokesperson, Raúl Lara, has described an 'enormous reduction' in workload for advertising and corporate voice-over. That is industry testimony, not official statistics, and it should be read as such. In the United States, SAG-AFTRA, the actors' union, ratified its new contract with the studios in 2026 with 91.4% of the vote. According to the union, it strengthens consent and pay for digital replicas, requires a business reason to scan an actor and protects minors. Press reports say it includes a safeguard so a performer's voice cannot be replicated in other languages without permission. Our reading: high-quality fiction dubbing is holding up better than catalog voice-over, tutorials or cheap advertising. In those areas, synthetic voice is already 'good enough'. It is the same pattern we see in other sectors: standardized work goes first, performance that requires judgment holds. But those who live on standardized work are hurting now, and that must be said plainly.

THE LEGAL MAP: NOT COPYRIGHT, BUT PERSONALITY RIGHTS. This is the technical point that causes the most confusion. Copyright protects a specific work: a recording, a fixed performance. It does not protect the sound of a voice. A federal court in New York made this clear in July 2025 in Lehrman and Sage v. Lovo. Two voice actors sued the company for cloning their voices. The judge dismissed almost all their copyright and trademark claims. But he let their contract, consumer protection and New York right-of-publicity claims go ahead. The right of publicity is the right to control commercial use of your own identity. That is the real battleground: personality rights. In the US, Tennessee's ELVIS Act, in force since July 2024, explicitly protects the voice. Its first major lawsuit came in November 2025, when Johnny Cash's estate sued Coca-Cola over an ad featuring a human sound-alike; the case is pending. At the federal level, the NO FAKES Act would create a right over digital replicas, with a notice-and-takedown process. It passed the Senate Judiciary Committee unanimously in June 2026. But on September 30, according to trade press reports, Senator Ted Cruz blocked a fast-track vote over free-speech concerns. In Europe, the AI Act has required labeling of deepfakes, including audio, since August 2, 2026, except for clearly artistic or satirical works. In Spain, Organic Law 1/1982 already names the voice: its Article 7.6 bars using it for advertising or commercial purposes without permission. In January 2026 the government approved a draft law to replace it, which explicitly covers AI-made 'ultra-impersonations'. It still has to pass Parliament. Denmark has gone further: its bill L 49, introduced on October 8, would protect face and voice through copyright law. And Mexico reformed its Federal Copyright Law in 2026 to require express consent and payment for AI use of voices. Our reading: the law already has the pieces. What is missing is making them work across borders. Serrano is protected in Spain, but the platform is in the US, where protection depends on the state.

TECHNOLOGY ALONE WILL NOT SAVE US. Watermarks are inaudible signals embedded in synthetic audio so it can be identified later. They help, but they are fragile. A 2025 academic study reviewed 22 audio watermarking systems and tested 109 attack configurations. It found that none withstood every attack. Other research shows that compressing, filtering or regenerating audio is often enough to erase a watermark. Detectors have a different limit: a negative result does not prove a recording is authentic. The real problem sits earlier, at the front door. In March 2025, Consumer Reports assessed six voice cloning services. Four of them, including ElevenLabs, relied on a checkbox in which users state that they have the right to clone the voice. Only Descript and Resemble AI required stronger proof, such as reading a consent statement live. Our reading: the effective defense is not catching fakes afterward. It is requiring verifiable consent beforehand. And at home, the cheapest safeguard is still a family code word for emergency calls.

IMPLICATIONS AND THE LONG VIEW. In the short term, the picture is tough. There will be fewer standard voice-over jobs, more phone scams and a cross-border legal gap that leaves people like Serrano fighting companies that do not even answer. We will not sugarcoat that. But the long-term direction is hopeful. For the first time, the law is starting to treat the voice as something each person owns and can license. That enables valuable uses: voice banks for people who will lose their voice to diseases such as ALS, affordable dubbing into minority languages, or an actor licensing a clone for minor jobs while keeping the performances they love, as Serrano himself suggested in an interview with ADSLZone. That is the abundance we argue for: machines doing repetitive work with permission and payment, and people choosing where to use their real voice. Getting there needs three things: verifiable consent by default on platforms, a voice right that can be enforced across borders and fast removal of unauthorized clones. Your voice is still yours. The law and the platforms now need to recognize it.