A voice can be cloned convincingly from about three seconds of audio. That fact, not a hypothetical, is what pushed a group of British actors to launch a legal campaign this week. It is also why the question worth asking about any voice cloning tool is not "does it sound real," every serious one does now, but "what happens before it clones anything at all."
Consent, done properly, is not a checkbox next to a wall of legal text. It is a specific sequence: someone reviews exactly what will be built from their voice, accepts a dated record of that decision, and only then does a sample reach a voice model. Most tools skip straight to the sample. Jellypod does not: cloning a voice, your own or a guest's, requires that acceptance to exist first, and the platform will not send audio to a voice provider without it.
What does consent actually mean for AI voice cloning?
Informed consent for voice cloning means the person whose voice is being used understood, before any recording was processed, what would be built, who would control it, and what it would be used for. That is a higher bar than agreeing to a platform's general terms of service when signing up. A blanket "I agree" clicked during account creation, months before anyone mentions cloning a voice, does not cover a specific decision to let that voice be reproduced.
One distinction matters more than any ethics checklist: consent versus licensing. Consent answers whether someone can clone a voice at all. Licensing answers what happens after: how long the clone can be used, in which projects, and whether it can be reused later for something the person never agreed to. A voice actor who consents to one campaign has not consented to every campaign that company runs for the next five years. Software that only stores a one-time yes-or-no answer cannot represent either of those distinctions, which is one reason a growing share of the debate has moved from ethics guidelines to statute.
Is AI voice cloning legal without consent?
Increasingly, no, and the legal ground shifted twice in the last two months alone.
In the United States, the NO FAKES Act cleared the Senate Judiciary Committee on June 18, 2026, on a unanimous voice vote. It would create the first federal property right in a person's own voice and visual likeness, licensable but not assignable, and it would not expire when the person dies. Before this, US protection against unauthorized voice cloning was a patchwork of state right-of-publicity laws that vary widely in what they cover and who can sue.
Denmark moved first. A copyright amendment proposed in June 2025 adds two new provisions, one protecting performers from unauthorized AI reproductions of a performance, one giving every person the right to demand removal of and compensation for unauthorized digital reproductions of their voice or likeness, with exceptions carved out for parody, satire, and criticism.
The UK does not have either yet, which is exactly what a new campaign is trying to change. Save Our Voices Now, backed by actors including Nicola Coughlan, Hugh Bonneville, and Matt Lucas, sent an open letter to Prime Minister Andy Burnham this week asking that voice ownership become a statutory right, the same legal footing as a person's name or image. Its founder, Peter Caulfield, put the urgency plainly: "In just three seconds, with the right AI system, your voice can be cloned, stolen and reinterpreted without your knowledge." The campaign backs that with a number: 28% of UK adults report having encountered a voice-cloning scam already.
| Jurisdiction | Status as of August 2026 | Mechanism |
|---|---|---|
| United States | Advanced out of Senate committee, not yet law | Federal property right in voice and likeness (NO FAKES Act) |
| Denmark | Proposed, expected to take effect late 2025 into 2026 | Copyright-style right, government takedown and compensation |
| United Kingdom | Campaign and petition stage, no bill yet | Requested statutory right, same standing as name or image |
None of that is theoretical for the people it is meant to stop. The FBI has publicly warned about scammers cloning a few seconds of someone's voice from social media, then calling a relative with a fabricated emergency and a demand for money sent by wire, gift card, or crypto. The Bureau puts total US losses to AI-enabled scams at more than $893 million in a single year. The law is only now catching up to a problem that has already cost people real money.
How do you get consent to clone someone else's voice?
The person whose voice it is has to be the one submitting the samples, not you uploading a recording of them on their behalf. That single rule rules out the most common shortcut: a producer grabbing an old interview clip, running it through a cloning tool, and calling it fine because the recording was public. Public availability of a recording is not the same as permission to clone the voice in it.
The practical version of this looks like an invitation, not an upload. On Jellypod, a producer who wants a colleague or a subject-matter expert as a recurring podcast host sends that person a single-use link. The guest opens it, no account required, records or uploads their own samples directly, and only their acceptance of the clone puts a new host in the producer's library. Nobody at the other end of that link can approve consent for someone else. This matters most in exactly the settings where a persuasive expert voice is valuable: a hospital's continuing medical education program inviting a physician to lend their voice to a recurring series, or a conference organizer turning a keynote into an ongoing show with the speaker's own voice attached to every episode.
What should a consent-first voice cloning tool require?
Four things, in order, before a single audio sample reaches a voice model.
First, the person reviews what will actually be built: the inferred name, the language, the backstory attached to the voice, not a generic paragraph of legal boilerplate. Second, accepting has to produce a real record, not a stored boolean. On Jellypod that record captures the exact policy text shown, its version, a timestamp, and who accepted it, tied permanently to the resulting voice. Third, that record has to exist before the provider gets anything: no tool should be sending samples to a cloning model while an approval step happens somewhere else in parallel, because a failure partway through then leaves a voice that was never actually authorized. Fourth, the record has to survive independent of the person who can see it in the product. An ordinary user should not be able to edit or delete their own consent history after the fact, because the whole point of the record is that it reflects what was agreed to at the time.
Most competing writeups on this topic stop at "get consent" and "be transparent." Few describe an actual sequence a tool enforces, because most were not built around one. A checkbox at signup is not evidence of anything. A timestamped, versioned record tied one-to-one to the voice it authorized is.
Frequently asked questions
Is it illegal to clone someone's voice without their permission?
It depends on the jurisdiction, but the direction of the law is toward yes. The US currently relies on state right-of-publicity statutes, which vary, while the federal NO FAKES Act would create a nationwide property right if it passes the full Senate. Denmark's copyright amendment already covers this more directly, and the UK is actively campaigning for similar protection.
Can I clone my own voice for a podcast or video?
Yes. Cloning your own voice raises none of the third-party consent questions above, since you are both the subject and the person accepting the terms. Jellypod's voice cloning lets you upload samples directly and review the inferred details before the clone is created.
What is the difference between consent and licensing for a voice clone?
Consent covers whether a voice can be cloned at all. Licensing covers the terms after that: how long the clone can be used, in what context, and whether it can be reused for something new. A person can consent to one use of their cloned voice without granting unlimited future use.
How do voice cloning scams typically work?
A scammer pulls a few seconds of someone's voice from a public video or social media post, clones it, then calls a family member claiming an emergency and asking for immediate payment. The FBI recommends hanging up and calling the person back on a known number before sending any money.
The short version
Consent for AI voice cloning is only real if it happens before the sample reaches a voice model, produces a permanent record, and comes directly from the person whose voice it is, never a third party uploading on their behalf. The law is moving toward requiring exactly that. See how Jellypod's guest invite flow puts that sequence into practice the next time you need a colleague, an expert, or a guest host cloned the right way.