Security
Last updated 24 September 2026
Envisianote holds your meetings and your work documents, so how they are protected matters more than most. This page describes the actual mechanisms — what is encrypted and with what, how your content is kept apart from other customers', what our AI providers may and may not do with it, and what happens when you delete something.
How your content is protected
In transit
Every connection between your device, Envisianote, and the services you connect is encrypted with TLS. We do not accept your content over an unencrypted connection.
At rest
Our PostgreSQL database and the Amazon S3 bucket holding recorded audio are encrypted at rest by the infrastructure providers.
At the application layer, on top of that
The most sensitive fields are encrypted again before they are written to the database, with AES-256-GCM, using a key held outside it. A copy of the database alone does not yield them. Specifically:
- the OAuth access and refresh tokens that let us reach Google, Slack, GitHub and Linear on your behalf;
- your meeting transcripts;
- the summaries generated from them;
- the name and email address on your account.
We list these rather than claiming everything is encrypted twice, because it is not. The decisions and action items we extract from your content are protected by the database and object-storage encryption described above, not by a second layer. We would rather be precise about where the line falls than round it in our favour.
Your content is kept apart from other customers’
Two mechanisms, at different layers:
- In the database. Every query is scoped to the owning account, so one customer’s records are not reachable from another’s session.
- In the search index. The vector index that makes your meetings and documents retrievable is a per-user namespace, created for your account alone — not a shared index partitioned by a filter. Your content is never pooled with another customer’s, and an answer to your question can only be drawn from your own namespace.
Your content is used to produce output for you, and is never shown to another Envisianote user.
AI processing, and training
We do not train our own models on your content, and we do not sell it. The providers that process your content to transcribe and generate — AssemblyAI, Anthropic and OpenAI — operate under API terms that prohibit training their models on content we submit.
Your content reaches a model only to produce something you asked for: a transcript of a meeting you recorded, a summary of it, the items in your daily briefing, or an answer to a question you typed. It is not used to improve the product for anyone else.
Data received from Google APIs carries a further, explicit restriction. Our use and transfer of it adheres to the Google API Services User Data Policy, including the Limited Use requirements: we do not use Google user data to develop, improve, or train generalised artificial intelligence or machine learning models.
Model output can be wrong, including when it sounds confident. We link back to the source material wherever we can so you can check it, and our Terms say plainly that briefings and answers are not a substitute for reading the source.
Deletion actually deletes
Deleting a note removes its transcript, its summary, and the recorded audio in S3.
Deleting your account removes your notes, transcripts, extracted items and stored audio, revokes the integration tokens we hold, and purges the vectors derived from your content out of the search index.
That last step is the one most easily got wrong, so it is worth saying how it works. The purge is driven by listing what is actually in your namespace, not by replaying our own record of what we put there — so it also removes anything our bookkeeping had lost track of. And a deletion that fails to purge those vectors is not reported as complete: the failure is recorded as an error rather than logged as a success. A deletion that left your content in a third-party index would not be a deletion, and we do not want to be able to tell ourselves otherwise.
Backups and log entries may persist for a short period after deletion before being overwritten on their normal cycle.
We ask for the narrowest access that works
Envisianote reads from services you connect, and nothing is connected by default. On the Google side in particular, the scopes are deliberately minimal:
- We cannot browse your Drive. Document access uses Google’s own file picker and the per-file
drive.filescope, which reaches only the documents you pick or that Envisianote itself creates. We cannot search, list, or index the rest of your Drive — not by policy, but because the permission does not exist. - Reading is read-only. Calendar and Docs are accessed with
calendar.readonlyanddocuments.readonly. We never create, modify, or delete your calendar events, and we never modify documents you created. - We hold no restricted scope. The broad
drive.readonlyscope was removed in August 2026 and is not coming back.
You can disconnect any integration at any time, which deletes the tokens we hold and stops all further access, or revoke access directly from your Google account permissions page.
Who else processes your data
We share your content only with the service providers that make the product work — transcription, model inference, hosting, the database, the search index, payments, error monitoring and analytics — each processing it on our instructions and for no other purpose. They are named individually, with what each one does, in our Privacy Policy. We do not sell your data or share it for advertising.
What we have not done yet
Envisianote is a small company shipping its first release, and a security page that lists only good news is not worth reading. So, plainly:
- We are not SOC 2 certified, and no audit is in progress.
- We have not commissioned a third-party penetration test.
- We do not yet offer single sign-on, audit logs, or customer-managed encryption keys.
We will update this page when any of that changes, rather than leaving you to infer it from silence. If one of them is a requirement for you, write to us and say so — it is useful to know what to build next.
Reporting a vulnerability
If you believe you have found a security issue, please write to privacy@envisianote.com with enough detail to reproduce it, and put “security” in the subject line so it is routed quickly. We will acknowledge your report, keep you updated while we investigate, and we will not pursue action against anyone who reports a genuine issue in good faith and gives us a reasonable chance to fix it before disclosing it.
Please do not run automated scans against production, access an account that is not yours, or degrade the service for other users while testing.
Contact
Security reports, privacy questions and data requests all reach us at privacy@envisianote.com. Our full disclosures are in the Privacy Policy and Terms of Service.