Security and governance

This page is the one we would want if we were the ones vetting a dictation tool: what actually happens to the audio, who can change that, and what we do and do not claim yet.

Where things stand today. On-device transcription, optional cleanup that is off until you switch it on, and no training on your dictation are true for every user. Teams can set one cleanup provider for everyone, described in step 03. We would rather tell you exactly what has shipped than round up.
01

Transcription happens on the device, always

Speech-to-text runs locally, using open, widely used speech models (Whisper, Parakeet and Moonshine) or, if you choose it on an Apple silicon Mac with macOS 26, Apple's built-in on-device recognizer. None of them is a black box on someone else's server. Audio is not a network request waiting to happen; it is processed where it was spoken, and Tulpa 1.0 does not save it.

02

Only the optional polish pass can send your words off the machine

Tulpa separates transcription, which is always local, from an optional cleanup pass that rewrites your dictation. Cleanup is off by default, and with Apple Intelligence (on Apple silicon) it runs on the Mac. Filler word removal and custom words run on the Mac with no AI. Your dictation leaves the device only through cleanup with a model that is not on your Mac. It sends text only, never audio. With a cloud model you connect with your own API key, after a warning you confirm, it goes straight from the Mac to that provider and never through us. On a Team plan it goes from the Mac to our server and on to the team's provider, and we do not store it. Checking for updates, downloading a speech model and, if you sign in, account and subscription checks reach the network too; none of them carries your audio or your transcripts.

03

Teams set one cleanup provider, and the key stays on the server

On a Team plan the administrator sets one cleanup provider for the team: an address, a model and an API key. The key is stored encrypted on our server and is never sent to any Mac. When a member uses team cleanup, the text goes from their Mac to our server and from there to the team's provider. We keep usage records (who, when, which model, token counts) and never the text. On Macs managed by an organisation, an MDM policy file can instead pin cleanup to local-only or to an endpoint the organisation names, and the app refuses to override it. Solo users choose for themselves.

04

Nothing trains on you by default

Tulpa does not use your audio or transcripts to train or improve any model. If cleanup is switched on with a cloud model, yours or your team's, text goes to that provider under its own terms, and Tulpa tells you so before you connect your own. That is a deliberate choice, not a default anyone opted you into.

05

Set once for a team, not seat by seat

A team's cleanup provider is set once in the admin's account and each member's Mac picks it up when they sign in, including new hires, with nothing for an admin to chase down machine by machine. A policy file on a managed Mac holds on that Mac whether or not anyone signs in. Tulpa ships for Macs with Apple silicon or Intel, running macOS 13.4 or later.

Does audio ever leave the device?

No. Raw audio is never transmitted, and Tulpa 1.0 does not save it; there is no code path that uploads it. Transcript text leaves only if cleanup is switched on with a model that is not on your Mac. With your own cloud key it goes straight from your Mac to that provider, under that provider's own terms. On a Team plan it goes through our server to your team's provider, and we keep usage counts, never the text.

Can one person change the team's cleanup setup for themselves?

On a Mac managed with an MDM policy file, the relevant settings are locked, and the app refuses the override outright rather than only hiding the control. The distinction matters, because a hidden setting can still be reached by anyone who knows where to look. A team's cleanup provider applies while a member is signed in to their seat, and signing out ends it. Cleanup itself stays off until someone switches it on.

What platforms does Tulpa run on today?

Macs with Apple silicon or Intel, running macOS 13.4 or later. Tulpa is a macOS app only: Windows and Linux are not supported.

Is Tulpa open source?

No. Tulpa is a proprietary commercial product. Its speech models are open and widely used (Whisper, Parakeet, Moonshine), so the core transcription behaviour is not a mystery even though the app itself is closed-source. On an Apple silicon Mac with macOS 26 you can also choose Apple Speech, Apple's closed recognizer, which also runs on the Mac.

Do you train on our audio or text?

No. The only way your text could reach a third-party model is if cleanup is switched on with a cloud provider you connect with your own key, or one your team administrator set, and what that provider may do with it is governed by its own terms. Tulpa warns you of that and asks you to confirm before it connects your own.

How is this different from a "privacy mode" toggle in other dictation apps?

A per-user toggle depends on every single person remembering to turn it on, it can usually be turned back off just as easily, there is nothing to audit, and nothing stops drift across a team. With Tulpa Teams, one cleanup provider is set by the administrator and used by every seat, and on managed Macs a policy file enforced at the app layer can hold cleanup to local-only, rather than leaving it to individual judgment.

Talk to us

Have a longer security questionnaire?

Send it over and we will walk through the architecture directly. Happy to go deeper than a landing page can.