What Popular Transcription Apps Actually Do With Your Recordings

Most people hit record on a transcription app without ever opening the privacy policy. Fair enough — they're long, dense, and written by lawyers. So we read a few of the popular ones, and looked at what's actually happening in the industry right now. Here's what we found.

Otter.ai is currently being sued over this exact issue

Otter.ai, one of the most widely used meeting transcription tools, is facing a federal class action in California. The core allegation: Otter's meeting bot joins calls and records them, but only the meeting host — not the other participants — is asked for consent. The lawsuit also claims those recordings are used to train Otter's speech recognition models (LegalClarity, Kilpatrick Townsend).

Otter's own privacy policy confirms the training practice, describing recordings as "de-identified" before being used to improve its models. The policy also lists several categories of outside vendors that receive user data, including cloud storage providers and third-party data-labeling services that help create training data (Otter.ai Privacy Policy). Otter's user-facing deletion flow moves a file to trash for 30 days before it's purged — but the policy separately reserves the right to keep backups and any data that already contributed to model training, which isn't undone by deleting the original recording.

It's not just Otter

Google Workspace's voice typing — the microphone icon in Google Docs — streams your voice straight to Google's servers for processing the moment you start dictating. Nothing is transcribed locally; it's all live cloud processing (analysis via Basil AI).

Rev, another major transcription service often used for legal depositions and interviews, discloses in its own privacy policy that it processes audio files — including deposition recordings — and collects a fairly broad range of personal data tied to your account and usage (Rev.com Privacy Policy). None of this means these companies are doing anything illegal — it's standard practice for a cloud-based service. It just means your recording is processed somewhere you don't control, by infrastructure and vendors you likely never checked.

The pattern

None of this happens because these companies are being careless. It's a structural consequence of how cloud-based transcription works: your audio has to leave your device and reach a server before it can be processed. Once it's there, it exists somewhere you don't control — which opens the door to things like model training, third-party vendors, and the kind of consent gaps at the center of the Otter lawsuit.

More from Offgrid Studio

Also read: Why Most Album Leaks Start in the Cloud, Not the Studio. If you work with audio in other ways, OffgridStem separates vocals, drums, and instruments from a track the same way — entirely on your iPhone.