Hark owl logo Hark Download free

Policy

Hark voice data policy

Biometric data retention and destruction policy. Last updated: October 30, 2026.

Hark keeps a voiceprint only of a person who signs a consent themselves, at the computer, with their own name: you, if you turn on Recognize my voice, and anyone else who chooses to set up their own voice there. Each voiceprint stays encrypted on that computer and External Software never receives it. Hark keeps no voiceprint of anyone else.

Who this policy is from

External Software LLC, Greenville, South Carolina, makes Hark, a desktop app that records and transcribes meetings on your computer. This is our written policy for voice data, including the retention schedule and destruction guidelines that biometric privacy laws such as the Illinois Biometric Information Privacy Act (BIPA) and the Texas Capture or Use of Biometric Identifier Act ask for. It covers the Hark app on every system (Windows, macOS and Linux). Our general privacy policy covers everything else.

What voice data exists

DataWhoseWhereHow long
A voiceprint (only for each person who signs the consent) That person's only Encrypted on that one computer. Never uploaded, synced or shared. At most 6 months (see below)
The recording of the person reading the consent passage That person's only On that computer Deleted as soon as the voiceprint is made
Voice analysis used to split a meeting into speakers Everyone in the recording On your computer, while the meeting is processed Discarded once the meeting is processed; never saved
Each consent record The person who signed External Software's servers. Contains no voice data. 6 years after the consent ends

Meeting recordings themselves stay on your computer. Hark never uploads audio to us, for transcription, sync or anything else.

The only purpose

Hark uses a voiceprint for one thing: to label that person as the speaker in the transcripts of meetings recorded with Hark on that computer. It is not used to verify your identity, not used to train artificial intelligence, and never sold, leased, traded or otherwise used to make money.

External Software never possesses voiceprints

Every voiceprint is made on the computer and stays there, encrypted, in Hark's own data. It is never sent to us, never included in sync (which is end-to-end encrypted and carries no audio or voice data), and never shared with anyone, including Anthropic, whose Claude models write Hark's AI notes from transcript text. We have no copy to disclose, sell or lose.

Retention schedule and destruction

Hark permanently deletes each voiceprint at the earliest of:

  1. 6 months after that person signs. They can then agree and record again.
  2. The moment the consent is withdrawn by choosing Delete my voice, or Delete next to the person's name.
  3. When a different External Software account signs in to Hark on that computer.

If you uninstall Hark, its data folder stays on the computer until you delete it. If you install Hark again, it deletes any voiceprint that is past its 6 months. Deleting Hark's data folder also deletes the voiceprints.

How to withdraw

You can withdraw at any time. In Hark on that computer, open Account and choose Delete my voice, or Delete next to your name under Voices Hark recognizes. Hark deletes the voiceprint straight away and tells us that you withdrew, so the consent record shows when it ended. If you set up your voice on someone else's computer and can't use it, ask them to delete it, or contact us: we'll mark your consent as withdrawn and ask the account holder to delete the voiceprint. Names already written in transcripts stay, because they are text. Voice recognition is optional: Hark works fully without it.

The consent records we keep

We keep a record of each consent as proof that it was given. It contains no voice data. It holds:

  • the email of the External Software account signed in to Hark on that computer,
  • the full name the person typed as their electronic signature,
  • the date and time they signed,
  • the full text of the consent they agreed to,
  • the text of the sentence they read aloud (the words, not the recording),
  • the IP address they agreed from, and the approximate location it indicates (country, region and city),
  • the name of the computer, the Hark version and the technical details your computer sends with the request (its "user agent"),
  • if the consent is withdrawn, when, and the IP address it was withdrawn from.

We keep each record for 6 years after the consent ends (when you withdraw, or when it expires), then delete it. Account holders can see the consent records made on their computers on their account page. We don't sell or share them, and disclose them only where the law requires it or to show that consent was given.

Other people's voices

Hark tells the voices in each recording apart only to split the transcript into speakers ("Speaker 1", "Speaker 2"...). That analysis happens on your computer and is discarded once the meeting is processed. Hark does not save voiceprints of the people you record and does not recognize them in later meetings, unless a person sat at your computer and signed the consent and read the passage themselves. Nobody can agree for someone else. You name the other speakers yourself, and those names are text in that meeting's transcript.

Recording consent is your responsibility. Many places require everyone's consent before a conversation is recorded; see recording consent laws by state. Our terms also require that each person sets up only their own voice, in person and of their own free will.

Changes to this policy

If we change how Hark handles voice data, we update this page and give the consent text a new version number. Questions: contact us.

Related

Frequently asked questions

Does Hark collect biometric data? +

External Software never collects or receives biometric data. If you choose Recognize my voice, Hark makes a voiceprint of your own voice and keeps it encrypted on your computer only. Other people can choose the same for themselves at your computer. Voiceprints are never uploaded or synced. We keep only a record of each consent, which contains no voice data.

Does Hark make voiceprints of the people I meet with? +

Not unless they choose it themselves. To split a transcript into speakers, Hark analyzes the voices in each recording on your computer and discards that analysis once the meeting is processed. Hark saves a voiceprint only of a person who sits at your computer, reads and signs the voice consent with their own name and reads a short passage aloud (Account, Add a person). Nobody can agree for someone else. Everyone else you name yourself, and Hark doesn't recognize them in later meetings.

How long does Hark keep a voiceprint? +

At most 6 months. Hark deletes each voiceprint at the earliest of 6 months after that person signed, the moment they withdraw (Delete), or when a different External Software account signs in to Hark on that computer. After 6 months the person can agree and record again.

How do I withdraw my consent? +

In Hark on that computer, open Account and choose Delete my voice, or Delete next to your name under Voices Hark recognizes. Hark deletes the voiceprint straight away and tells us, so the consent record shows when it ended. If you set up your voice on someone else's computer and can't use it, ask them to delete it, or contact us and we'll mark your consent as withdrawn and ask them to. Names already written in transcripts stay, because they are text.

Does sync copy my voiceprint to my other computers? +

No. Your voiceprint never leaves the computer it was made on, even with sync turned on. On another computer you agree and record again. Sync is end-to-end encrypted and carries transcripts, notes, screenshots, AI notes and settings, never audio or voice data.