Subprocessors
Mandra Voice
Version 1.1 · Last updated 27 September 2026
To run Mandra Voice we use a small number of service providers ("subprocessors") that process personal information on our behalf. Every one of them is bound by a written agreement requiring them to process personal information only on our instructions, to keep it confidential and secure, to help us respond to your privacy rights requests, and to delete or return it when our relationship ends.
Notice of changes. We will publish and email notice at least 30 days before adding a new subprocessor. If you object to an addition, you may cancel your subscription before it takes effect and we will refund the unused portion of your term.
To be notified of changes, email support@mandravoice.com with the subject "Subprocessor notifications".
Infrastructure
| Subprocessor | What it does | Personal data involved | Location | Transfer mechanism |
|---|---|---|---|---|
| Supabase, Inc. | Database, authentication, file storage, backend functions | Account details, transcripts, settings, usage, device sessions, profile pictures | Canada (AWS ca-central-1) | Data stays in Canada; EEA/UK transfers covered by Canadian adequacy |
| Neon, Inc. | Standby database, used only during a failover from the primary | Same as Supabase, while a failover is active | United States (AWS us-east-1) | SCCs |
| Vercel, Inc. | Hosting for the website, admin portal, and payment webhook | IP address and request metadata; billing webhook payloads | Global edge network | EU–US Data Privacy Framework |
| GitHub, Inc. (Microsoft) | Distributing the application and its automatic updates; storing our encrypted nightly database backups (kept 30 days) | IP address and app version when checking for updates; encrypted backup copies of account data, unreadable without our key | United States | EU–US Data Privacy Framework |
Speech and AI processing
| Subprocessor | What it does | Personal data involved | Location | Transfer mechanism |
|---|---|---|---|---|
| OpenAI, L.L.C. | Speech-to-text (Whisper), and AI text improvement | Voice audio (not stored by us); transcript text | United States | API terms: no training on customer content; abuse-monitoring retention up to 30 days |
| Deepgram, Inc. | Speech-to-text | Voice audio (not stored by us) | United States | API terms: no training on customer content |
| ElevenLabs, Inc. | Speech-to-text fallback, and Read Aloud voices | Voice audio (transient); text to be spoken aloud | United States | EU–US Data Privacy Framework; logging explicitly disabled per request |
| Google LLC (Google Cloud / Gemini) | AI text improvement, when selected as the active provider | Transcript text | United States | EU–US Data Privacy Framework |
| Groq, Inc. | Speech-to-text (Whisper, the default engine), and AI text improvement when selected as the active provider | Audio recordings of dictation; transcript text | United States | SCCs |
No provider may train on your data, and none stores your audio as a matter of course. We
use each provider under its published API and data-processing terms, which prohibit using
customer content to train models. Those terms are not the same as a negotiated zero-retention
agreement: a provider may hold content briefly for abuse monitoring — OpenAI, for example,
retains API inputs and outputs for up to 30 days unless zero retention is enabled on the
account. Where a provider defaults to logging we disable it per request; for ElevenLabs that
is the enable_logging=false flag, which opts out of their default retention and training
pathway.
AI text improvement routes to one of OpenAI, Google, or Groq depending on which provider is active for the Service at the time. We list all three because any of them may handle your text. Audio is never sent to these providers — only the text produced from it — and only if you have AI enhancement enabled, which you can turn off in Settings.
EU and UK data residency. When Deepgram is the active speech engine, audio from European
time zones is routed to Deepgram's EU endpoint (api.eu.deepgram.com) so it is processed
within the EEA. This applies to Deepgram only; the other engines and all AI text features
process in the United States.
Billing, communications and diagnostics
| Subprocessor | What it does | Personal data involved | Location | Transfer mechanism |
|---|---|---|---|---|
| Stripe, Inc. | Payments, subscriptions, customer portal, fraud prevention | Name, email, billing address, payment method (Stripe holds the card; we see only brand and last four digits), transaction history | United States and EU | EU–US Data Privacy Framework; SCCs |
| Resend, Inc. | Sending transactional and notification email | Email address, name, message content | United States | SCCs |
| Functional Software, Inc. (Sentry) | Crash and error reporting | Error details, stack traces, app version, operating system, account identifier; scrubbed before sending | United States | EU–US Data Privacy Framework |
What is not on this list
- We use no advertising networks, no analytics vendors, and no data brokers. There are none to list.
- We do not sell or share personal information with anyone, for any purpose.
- Our own contractors and employees are not subprocessors; they are bound by confidentiality obligations and access personal information only where necessary to support you.
Transfers out of Canada
We are a Canadian organisation and your account data, transcripts, and settings are stored in Canada (AWS Canada Central). The speech, AI, payment, email, and diagnostics providers above process data in the United States. The European Commission has determined that Canada offers an adequate level of protection for personal data transferred to commercial organisations subject to PIPEDA (Decision 2002/2/EC), which covers transfers from the EEA to us. Onward transfers to United States providers rely on the EU–US Data Privacy Framework where the provider is certified, and otherwise on Standard Contractual Clauses with supplementary technical measures — encryption in transit and at rest, and the providers' contractual ban on using customer content for model training.
Quebec residents: we assess privacy risk before transferring personal information outside Quebec, as Law 25 requires.
You may request a copy of the transfer mechanism for any provider at support@mandravoice.com.