Are meeting recordings private?
Mostly, no. Recordings made by cloud notetakers live in the vendor's cloud under the vendor's retention rules, and your employer can typically access anything in a company-managed account. Meeting platforms keep recordings tied to the host's account. To actually keep them private: record without a bot joining the call, keep raw audio in a local folder you control, and process it on servers you can hold accountable.
The default state: not private
Most recordings are made by tools whose default architecture puts your audio in someone else's cloud:
- Cloud notetakers (Otter, Fireflies, Fathom) - the bot records, the audio and transcript live in the vendor's storage under the vendor's retention rules.
- Meeting platforms (Zoom, Teams, Meet) - cloud recordings are tied to the host's account, often shared by link, and governed by the host's organization policies.
- Local recordings (QuickTime, Voice Memos) - the audio stays on your machine, which is private by default, but you get a raw file with no transcript and no structure.
The first two trade privacy for convenience. The third trades convenience for privacy. The interesting question is: can you get both?
Who can actually see what
Privacy is about access. Map each recording type to who can reach it:
| Recording type | Who can see it | Retention | Deletion |
|---|---|---|---|
| Cloud notetaker (Otter, Fireflies) | You, your team workspace, sometimes the vendor for abuse review | Vendor's retention rules | Per the vendor's policy |
| Zoom cloud recording | Host's account; shared links (often forwarded beyond the intended audience) | Per host organization's settings | Host controls |
| Local file on your Mac | You (and anyone with your disk access) | Yours - as long as you keep it | Yours - delete when you decide |
The pattern is clear: the further from your machine, the less you control who sees it and how long it's kept.
Three rules for private recordings
- Record without a bot joining the call. A notetaker bot is a third party in the room - visible to clients, sometimes blocked by security policies. A recorder that captures system audio from your Mac leaves the meeting untouched. See how to record without a bot.
- Keep raw audio in a local folder you control. Not a vendor's cloud, not an expiring link. A plain folder you can back up, move, or delete on your own schedule - see how to keep files local.
- Process locally where you can. Transcription on your own machine means your audio never reaches a third party; where a server is unavoidable, a plain privacy policy beats "we use third-party AI services, see their terms".
Nerpa was built around these three rules: it records calls without a bot joining, keeps raw audio in a local folder on your Mac, and transcribes on your computer. The result: a speaker-aware transcript with decisions and action items, organized by person - without a vendor's copy of your audio.
Rule two has a strong follow-through: once audio is a local file, disk encryption makes it unreadable to anyone but you. The practical steps are in the encrypted voice recorder guide.
Private by architecture, not by promise
No bot in the call. Raw audio in a local folder. Transcription on your computer. Free to try.
Latest stable release. Requires macOS 12+
How Nerpa handles privacy
The architecture is stated plainly, so you can verify rather than trust:
- Recording: the Mac app captures system audio via Apple's Core Audio Process Tap - the meeting itself is never touched, no bot joins, and the audio goes to local storage on your Mac.
- Storage: raw audio lives in a plain folder on your Mac (
nerpa_storage), named by date, standard format - playable, movable, deletable by you. - Transcription: runs on your computer - no third-party cloud AI service and no upload of your audio.
- Result: speaker-aware transcript with a summary, decisions, and action items, organized by person - the client history that builds itself.
Honest limits: transcription stays on your computer, but Nerpa is not a zero-knowledge vault - AI features (notes, summaries, questions) run on plan credits, and device sync goes through EU-hosted servers. We don't claim more than that. What you get is a clear, stated architecture instead of a vague privacy page.
The honest limits
Privacy isn't binary, and no tool is perfect. What Nerpa doesn't do:
- No built-in encryption of raw audio. Raw audio lives as a plain file on your Mac, so its protection is your Mac's: FileVault, a strong password, an encrypted disk image.
- No self-hosted server. Transcription runs in the app on your computer; AI features and sync use Nerpa's servers, and there is no on-premise server edition.
- Sync data lives in the account. If you use sync and share links, those follow standard account rules - delete the account, delete the links.
For most users this architecture is a strict upgrade over cloud notetakers. For organizations with strict data-sovereignty requirements, a fully local setup (Whisper + manual workflow) remains the alternative.
Frequently asked questions
Are meeting recordings private by default?
No. Cloud notetaker recordings live in the vendor's cloud under the vendor's retention rules, and meeting platform recordings are tied to the host's account. Local recordings on your own machine are the exception - private by default.
Who can see my Zoom recordings?
Anyone on the host's account with cloud-recording access, plus anyone the host shares a link with - and shared links get forwarded. Recordings sit on the platform's storage under the organization's retention policies.
Can my employer see my meeting recordings?
For recordings in company-managed accounts (Zoom, Teams, Otter Business): typically yes - IT and compliance teams have access, and work-related communications may be reviewable under company policy or law. Local recordings on your own machine are yours unless policy says otherwise.
How do I keep my meeting recordings private?
Three rules: record without a bot joining the call, keep raw audio in a local folder you control, and process on servers you can hold accountable. Nerpa does all three - see how it fits a client-call workflow.