Trust & safety
Powerful agents, with safety built in
Agents that act on their own, and realistic voices, must be hard to misuse. These are the protections built into Lipivo.
Voice cloning needs the speaker's consent
- The speaker reads a consent sentence aloud, with two code words that change every time, so an old or borrowed recording cannot be used.
- The recording is transcribed and checked against the sentence before any voice is made.
- The consent record is kept with the voice, and a cloned voice can only be used by the account that made it.
Built to stop voice scams
- Agent prompts, procedures and every agent reply are checked against known voice-scam scripts: OTP, PIN or CVV requests, KYC threats, 'digital arrest', remote-access apps and lottery fees.
- Text sent to the speech API is checked too; refused requests count towards automatic suspension of the account.
- Anyone can report misuse of a generated voice; the request id in the audio links a report to the account.
Outbound calls that follow the rules
- Calls are placed only between 9 am and 9 pm India time (TRAI calling hours).
- Numbers on your do-not-call list are never rung, including numbers added after a campaign started.
- Calls go out from your own number with your own phone provider.
Agents stay under your control
- Guardrails limit the topics an agent covers, block words you choose and hide personal data in its replies.
- Sensitive actions, such as refunds or account changes, can wait for a person's approval before they run.
- Your team can watch any live conversation and take it over at any moment.
Data we keep, and data we don't
- Call recordings and generated audio are not stored. Audio played on a call is held in memory for a few minutes only.
- Conversation logs are kept for 90 days, and you can delete any of them earlier.
- Hosted on AWS (Mumbai region), with daily database backups.
Keys and secrets
- Your phone provider and AI keys are stored encrypted; the dashboard never shows them again.
- API keys are stored only as a hash, can be revoked at any time, and can have a monthly spending limit.
- Webhooks we send are signed, and webhooks your agents call cannot reach private networks.
Teams and accountability
- Teammates sign in with their own accounts; owners and admins decide who has access, and removing someone ends their access at once.
- An activity log records who changed what in a workspace, without storing prompts or keys.
Heard a voice you think was misused?
Tell us; you do not need an account.