Biometric policy

What it is

Partner mode is for two people in one room: you, and someone asking you questions. It is the "In person" option when you set up a mock. One phone on the table hears you both, and the room needs to know who said what.

To do that, it turns stretches of speech into a set of numbers that describe a voice, called a voiceprint, and compares each new stretch with the voiceprints it has. Only partner mode does this. A mock with the AI interviewer doesn't.

How a voiceprint is made

• You: a short voice check at the start of the mock.
• Your partner: their first question or, for the preview, a short phrase they read out loud.

Each voiceprint can be refined by the speech that follows, for that mock only. A voice that matches neither of you is left out of the mock.

Where it is held

On our server, in memory, for that one mock. A voiceprint is never written to a disk, a database or a log. It is never sent to your phone, to our speech services or to anyone else. Matching runs on our own server, not at another company.

When it is destroyed

The voiceprints are dropped when the mock ends or you leave it, after 20 minutes with nobody in the room, or if the server restarts, whichever comes first. Nothing is kept for a later mock: every partner-mode mock starts afresh.

What is kept

No voiceprint is kept: not yours, and not your partner's. What the mock's record does keep, for each stretch of speech from either of you, is:

• the words, as text;
• who the room thinks said them;
• a match score against your voice and, once your partner's voice is known, one against theirs, which we keep to tune the matching;
• when it was said, how long it lasted and how loud it was.

A match score is one number saying how close two voices were. It isn't a voiceprint, and it holds no sound.

This record belongs to the account holder, and it is in their export. It stays until they delete their account, or until the account has gone unused for 12 months, and copies in our backups age out within 5 weeks after that. The sound itself goes to a speech service to become text, and we don't store it.

Both people must agree

If you run a partner-mode mock, you must tell the other person, before you start, that their voice will be matched this way, and that their words and a match score for each stretch of their speech will be stored in your account. You must get their agreement. Show them this page. If they don't agree, or you aren't sure, use the AI interviewer instead.

Partner mode is only for two people who are in the same room. It isn't for capturing a call, a meeting or a real interview.

What we don't do

We don't sell, lend or share voiceprints. We don't use them to work out who someone is, only to tell the two people in that one mock apart. We don't train any AI model on them.

Questions

A partner who has no account can write to us about a mock they were in. The account holder can delete that record at any time by deleting their account.

Contact

EZ Life Solutions Corp.

Email: hello@ezlifesolutions.ca.