The 2026 baseline
Cloning is cheap and can’t be stopped at the source. What enrolment gives you is a check: verify any clip against your voiceprint in about 90 seconds.
In 2023 the FTC put out a public advisory about voice-clone family-emergency scam calls. By 2025 the calls were measured in millions. Per published voice-cloning research, the audio needed to seed a clone has dropped to a few seconds of conversational speech. That’s well within reach of anyone who can find a voicemail greeting, a podcast appearance, or a recorded webinar.
Three seconds is enough to clone your voice, and enrolment doesn’t stop that. Nothing stops cloning at the source. What enrolment gives you is a check. Any clip can be verified against your enrolled voiceprint in about 90 seconds. So you can tell whether it is really you.
How storage works, in plain language
Enrolment stores a non-reversible voiceprint under EU residency, raw audio deleted at session close, and you can one-click delete all of it.
You schedule a forty-minute session. You read a documented consent script and a fixed phrase set. The recording is processed at once into a non-reversible voiceprint. That’s a numerical signature of your voice. It’s enough to run a similarity check, but not enough to rebuild your original audio.
The voiceprint and consent record are stored on our own infrastructure. The raw audio is deleted at session close. We keep the voiceprint, the consent record, and session metadata. You can delete all three at any time from your dashboard. It takes one click, with a 24-hour confirmation. We default to deletion on downgrade or account closure, because holding biometric identifiers past need is the wrong posture.
The legal framing is on the biometric policy page: BIPA-aligned consent, bounded retention, no sale or transfer, documented destruction. We hold to that standard globally, even where local law is lighter.
What you actually do with an enrolled voiceprint
Verify a suspicious clip on demand. On Executive, daily background scans check public audio channels for clips that match your voice.
On-demand verification is the main use. A relative gets a panicked voice call and records the next one. You upload the recording, and we compare it against your enrolled voiceprint. You get a verdict in about 90 seconds: match, no match, or inconclusive with an explanation. That’s the family-scam scenario.
On Executive, once you’re enrolled, daily background scanning covers public audio channels. That means podcasts, video platforms, and the dump channels where impersonating clips typically surface. When a clip you didn’t author matches your voiceprint, you get an alert. Pair with deepfake verification for face-plus-voice attestation when a case needs both signals.
Who shouldn’t enrol
If your threat model is “I don’t want any company to hold any biometric data about me, ever” — don’t enrol. That’s a coherent position and we respect it. Voiceprint protection trades one narrowly stored, non-reversible biometric artefact for the ability to verify clips against a baseline. If you’d rather not make that trade, the rest of the Executive tier still works without it.
See also
- Biometric policy — BIPA-aligned consent and retention.
- Deepfake Verification — verdicts against enrolled voiceprint.
- Executive Identity Defense — the tier voiceprint ships under.
- Pricing — Executive and Team tiers.