Handling personal data
Email is personal data almost by definition: an address identifies a person, and the body usually says something about them. This page is about what crosses the API and what happens to it afterwards.
What a send stores
A message you submit is stored, in full, including the body:
| Recipients | To, Cc and Bcc |
| Subject and body | HTML and text, as rendered |
| Attachments | inline in the request, stored with the message |
| Template variables | the values you substituted |
| Delivery outcome | status, SMTP response, bounce reason |
That is what makes GET /v1/messages/{id}, the sandbox inspection and bounce
diagnosis possible. It also means anything you put in a send is retained,
including anything you put in a template variable.
Data minimisation, concretely
Three habits that cost nothing:
Send identifiers, not records. order_id plus a link beats embedding the
order. The link is access-controlled by you; the email is not, because email
is forwarded, printed, synced to phones and left in inboxes for years.
Do not use Bcc as a mailing list. Every Bcc recipient is stored on the message, and a bug that renders recipients into a body discloses all of them at once. Use batch sending, which gives each recipient their own message and their own delivery result.
Keep diagnostics out of the subject. Subjects are logged by every mail server on the path, ours included, and they appear in notification previews on lock screens.
What we keep, and for how long
Messages are retained so you can inspect and diagnose them. A tenant retention policy controls mailbox contents; the sending record is ours.
Suppression entries persist deliberately, because their purpose is to outlive the message that created them. A hard bounce or a complaint that expired would be re-sent to, which is the behaviour suppression exists to prevent. See Suppression management.
Delivery events keep the recipient, status and SMTP response.
If you need a specific retention commitment in writing, that belongs in the DPA rather than here, and the answer may differ per contract.
Deletion requests
A person asking you to delete their data is your obligation as controller, and we are the processor. Two things are worth knowing before you promise a timeline:
Removing a suppression entry re-enables sending to that address. If somebody complained and you delete the record of it, the protection goes with it. Usually the right answer is to keep the suppression and delete elsewhere, and that is a defensible position: the entry exists to honour their earlier objection.
A sent message is in the recipient’s mailbox too, which is outside both your control and ours.
Encryption
In transit, TLS 1.2 or 1.3 on every API and mail connection, with forward secrecy. See TLS requirements.
At rest, be precise about what is and is not covered, because the difference matters:
| Migration credentials | AES-256-GCM, encrypted before storage |
| API key secrets | hashed, never stored recoverably |
| Mailbox passwords | argon2id |
| Mailbox contents, with the E2EE add-on | ciphertext the server cannot read |
| Mailbox contents, without it | stored as the message, protected by access control rather than by encryption |
End-to-end encryption is a separate, opt-in add-on available on every plan. Where it is enabled, keys live on the user’s device and the server holds ciphertext it cannot read. That is the only configuration in which mail content is encrypted at rest in a way that survives access to the server.
It changes what the platform can do, and the trade is deliberate: an E2E mailbox cannot be read by the server, so agent access is refused for it and server-side features that need the plaintext do not apply.
Mail to and from external providers cannot be end-to-end encrypted by any provider, which is a property of SMTP rather than of this platform.