Trust and safety

1. How consent works

A cloned voice goes live only after the person it belongs to has agreed, checked in one of two ways.

A recorded consent phrase. The speaker reads a short statement in their own voice, naming themselves and the customer, with three random code words:

“I, Priya Sharma, am the owner of this voice. I consent to Sunrise Academy creating and using a synthetic copy of my voice. My verification code is mango, rocket, violin.”

We check that every code word was spoken, so the recording was made for this request and is not an old clip played back, and that the voice in it matches the reference recordings. The speaker has five attempts within 24 hours.

A signed release. When consent was collected another way, for example a release form signed by a teacher, our team records the release, where it is kept and who checked it, against the voice before it goes live.

Either way, every voice keeps a record of how consent was given, when, and the result of each check. A voice's reference recording can only be replaced by a recording of the same person.

2. Voices we will not clone

• Anyone who has not agreed themselves, including a colleague, a family member or a former employee.
• Public figures, celebrities or election candidates, without their written permission.
• Children, without consent from a parent or lawful guardian.
• A voice copied from a video, call, podcast or social media post, when the speaker has not agreed to it.

3. What Dhvni must not be used for

• Impersonating a real person, or misleading listeners about who is speaking.
• Fraud, such as fake calls from a bank, a government office or a family member, or collecting OTPs and payments under false pretences.
• Making a real person appear to say something they did not say.
• Harassment, threats or abuse of anyone.

Where the law requires it, and in calls to people who have not chosen to talk to an AI, listeners must be told they are hearing a synthetic voice. The full rules are in our acceptable use terms.

4. Report a misused voice

If you think a voice made with Dhvni is being misused, or your own voice was cloned without your agreement, write to support@dhvni.in. You do not need to be a Dhvni customer. Please include:

• The audio, or a link to where you heard it.
• Where and when you heard it, for example a phone call, an app or a website.
• Whose voice you think it is, and whether it is yours.

5. What happens after a report

• We reply to confirm we have your report.
• When a report points to a voice on Dhvni, we disable that voice while we review, so it cannot be used.
• We tell the customer whose account the voice is on, and ask for their consent record.
• If consent was not valid, we delete the voice and its reference recordings, and may close the account.
• We tell you the outcome, within 30 days of your report.

6. How we handle data

• Audio sent for transcription is not stored after we return the text.
• Text sent for speech is not stored. We keep only usage counts, such as characters spoken.
• Reference recordings and trained voices are deleted when the voice is deleted.
• Consent records are kept for three years after a voice is deleted, so we can show who agreed and when.
• Security logs are kept for 180 days.

The details, and your rights over your data, are in our privacy policy.

7. Security

• All traffic to the API is encrypted in transit with TLS.
• API keys are stored only as a hash and shown once, when they are created.
• Browser tokens last two minutes and open only the speech stream, so a full key never has to reach a browser.
• Access to customer data is limited to the people who need it to run the service.

8. Providers that process data for us

The providers that help us run Dhvni, what each one does and where, are listed in our privacy policy. We update that list before a new provider starts processing customer data.