Legal
Sub-processors
Version 2.1 · Last updated: 2 November 2026
These are the providers that may process session content on our behalf. This list forms part of the Data Processing Terms: its Sub-processors provision says how the list changes, how much notice you get, and how to object. Every provider listed is authorised to receive session content; which of them serves a given session is our configuration choice.
This is the version of 2 November 2026. What a provider does with what it receives, and how long it keeps it, is set by that provider’s own terms and can change; the entries below are as we last checked those terms, on 5 October 2026.
| Provider | We send it | What it does with it | It keeps it for |
|---|---|---|---|
| Gladia SAS France (EEA) | Audio | Speech recognition and translation. Returns the transcript and its translations. It does not use what it receives to train its own models. It gives us a way to pass a deletion request on to it. | Up to 1 day |
| ElevenLabs, Inc. United States | Audio | Speech recognition. Returns the transcript. It does not use what it receives to train its own models. It gives us no way to pass a deletion request on to it. We will state the period its terms set here before it processes any session. Its use of what we send it for training was switched off in our account on 23 October 2026. | Its own terms set the period |
| Google LLC Processed globally — no regional endpoint | Text | Machine translation of the transcript. Returns the translations. It does not use what it receives to train its own models. It gives us no way to pass a deletion request on to it. Authorised Google staff may review what it keeps, for that purpose. | Up to 55 days, to detect abuse |
| Anthropic, PBC United States | Text | Machine translation of the transcript. Returns the translations. It does not use what it receives to train its own models. It gives us no way to pass a deletion request on to it. | Up to 30 days |
| Microsoft Corporation United Kingdom; its translation service is called on a global endpoint | Text and audio | Hosts the Service, its databases and its session store, may provide machine translation, and relays the transcript files a guest has asked for. It does not use what it receives to train its own models. It gives us no way to pass a deletion request on to it. | Not kept for its own purposes |
Providers we use only for data where WEYD is the controller, such as our email and support tooling, are not sub-processors of session content and are not listed here; the Privacy Policy covers those. Where a provider above is outside the United Kingdom, International transfers, of the Data Processing Terms states what we rely on for the transfer.
Review and changes
We review this list, and each provider’s own terms, at least once a year and whenever we intend to change it. A change to it follows Sub-processors, of the Data Processing Terms. Every version from 2 November 2026 stays published at its own address, so a list you accepted remains readable after it is superseded.
There is no earlier published version.
Questions and objections: hello@parlabase.com.
Version 2.1 · Last updated 2 November 2026
- Version 2.1, 2 November 2026 (this version) — Each version from this one is published at its own address