AI Privacy & Consent
When you sign up, you agree to let MailPrism use AI providers (Google Gemini as the primary provider, with OpenAI as the fallback) to analyze your email. You can turn AI off at any time in Settings → Privacy → AI Data Processing. While it's off, nothing is sent for AI analysis.
Consent is one half; your plan is the other. AI analysis is included on paid plans (Starter and up). The Free plan runs rule-based automation, though it can still use the AI rule drafter — 10 AI drafts a day. Bring-your-own-key (BYOK) is available on Business, Enterprise, and Lifetime.
How consent works
AI processing is controlled by a single consent setting. While it's off:
- No email content is sent to an AI provider for analysis.
- All AI features stay disabled — email classification, smart suggestions, and any rule conditions that depend on AI.
- Your non-AI rules keep running normally on their plain conditions.
While it's on, MailPrism may analyze your emails with AI to power those features.
The consent setting is the master switch. When it's off, every AI feature is off — there's no way for AI to run without it.
Where to turn it on or off
Consent starts on: the sign-up form asks you to agree to AI processing, and you can't create an account without ticking that box. After that, there are two places consent appears:
- Settings → Privacy — the AI Data Processing card has a toggle labeled AI Data Processing Consent. A badge shows whether it's Active or Disabled, and the date you consented.
- The dashboard banner — if you've turned AI off, the dashboard shows an Enable AI-powered features prompt with Enable AI Features and No Thanks buttons.
To turn AI on, flip the toggle on (or click Enable AI Features on the banner).
To turn it off, flip the toggle off. MailPrism asks you to confirm, because this immediately disables all AI features — including email classification, smart suggestions, and rule automation. You can re-enable at any time.
What's sent for analysis vs. what's stored
It helps to separate the two:
| What happens | |
|---|---|
| Sent for analysis | The email's subject, sender, recipients (To and Cc), and body text. For reply tracking, MailPrism also sends the thread — each message's sender, recipients, date, and text — plus your own email address, so the AI can tell which messages are yours. |
| Stored in your account | See the list below. |
What MailPrism stores
| What | How long |
|---|---|
| A cache of recent messages | To display your inbox, MailPrism keeps an encrypted (AES-256) copy of the messages it processes. Message bodies are deleted after 7 days without being opened, and fetched again from Gmail when you open the message. Details such as sender, subject, date, and labels are kept until you delete your account. |
| AI results | The signals and summaries AI produces are kept with your rule and action history, so rules can use them and you can see why something happened. |
| Sent mail you asked MailPrism to learn from | If you use Learn from your sent mail, the sent emails it analyzes are stored for that writing profile until you press Forget what was learned. |
Your mailbox itself stays in Gmail.
MailPrism uses the providers' paid API services. Under those terms, your email is not used to train their models. Providers may keep requests for a limited time to detect abuse — the Privacy Policy lists each provider's retention period.
A few details worth knowing:
- MailPrism only analyzes an email when there's a reason to — many routine, machine-generated emails are handled without sending anything to AI.
- Identical emails can reuse a previous result instead of being re-analyzed.
- Your Gmail access is always stored encrypted.
For the providers behind analysis and how to use your own, see Bring your own key (BYOK).
What turning AI off does to your rules
Turning AI off doesn't break your automations. Rules that use AI conditions (like urgency is high) simply stop matching on those conditions. Every rule built on plain conditions — sender, subject, labels, timing — keeps working exactly as before.
For important automations, pair an AI signal with a concrete condition so the rule still does something sensible even if AI is ever turned off. See Best practices.
Related
- AI, explained — what the AI reads and how it's used.
- AI usage & cost — track operations, tokens, and spend.
- Privacy & Security — the broader picture.