How to protect your voice from being cloned

Start from an honest premise: if you have posted video or audio publicly, there is enough of your voice online to clone it. Modern systems need seconds. No amount of care undoes what is already published, and pretending otherwise would be dishonest.

So the goal is not prevention, which is not achievable. It is reducing exposure, making misuse detectable, and removing the leverage that voice fraud depends on.

The fraud actually works like this

Voice cloning fraud is rarely sophisticated. The common pattern is a short, urgent call: a family member in trouble, a senior colleague needing an immediate transfer, a distress call late at night. The cloned voice does not need to be perfect, because the target is not listening critically — urgency is doing the work.

That is the weakness to attack. Fraud that relies on urgency fails against a household or a finance team that has a verification habit.

What actually helps

Agree a verification step with people who matter

A shared word or a call-back rule with family, and a mandatory second-channel confirmation for any payment instruction at work. This is the single most effective control, it costs nothing, and it defeats a perfect clone as easily as a poor one.

Never act on voice alone for anything financial

Treat an inbound voice as an unverified identity claim, the same way you already treat an email address. Hang up and call the number you already have.

Reduce unnecessary public audio, without becoming a hermit

Lock down personal accounts that carry voice notes and stories. If you are a creator, your public audio is your job — but there is no reason your private circle's content should be public too.

Be careful with cold callers who keep you talking

Extended unsolicited calls that prompt you to say specific words are a recognised harvesting pattern. You do not owe an unknown caller a conversation.

How to spot a cloned voice

Detection by ear is unreliable and getting worse, so treat these as weak signals rather than tests:

• Flat or wrong emotion under circumstances that should produce strong emotion. Clones read; they do not feel.
• No overlap or interruption. Real conversation is messy. Generated speech does not talk over you or lose its thread.
• Unnaturally even pacing, no breath noise, no self-correction.
• Resistance to being interrupted with a question only the real person could answer. This one is the reliable one — it is not a detection technique, it is verification.

If someone has cloned your voice

1. Preserve evidence. Save the audio, the URL, timestamps and any account details before reporting — content disappears once reported.
2. Report to the platform hosting it. Every major platform has an impersonation route, usually faster than any other option.
3. Report to the service that generated it, if you can identify it. Responsible providers have a grievance route and can act on the account. Mayantar's is in the footer of every page.
4. File a complaint with the relevant cybercrime authority, particularly if money was involved.
5. Warn your circle directly. A short message to family, colleagues or your audience removes the fraud's leverage immediately, which matters more than any takedown timeline.
6. Take legal advice if it is commercial or persistent. Personality rights in India cover voice, and courts have granted protective orders.

What platforms should be doing

Worth knowing so you can judge any provider, including us: consent captured and stored at the point of cloning; submitted reference clips screened against known voices, with matches held for human review rather than auto-approved; watermarking retained on generated audio rather than stripped; a monitored abuse address and a real takedown process; and clear retention and deletion terms.

What we do specifically is on the safety and consent page. A provider that cannot answer these questions plainly is telling you something.