Privacy
Hackathon build, last updated November 4, 2026. Contact: the person who invited you to Prudence.
What Prudence does with your Gmail
With your permission (Google scope gmail.modify), Prudence reads new messages in your inbox to judge whether they are scams, adds a "Prudence" label to suspicious ones, and moves likely scams out of your inbox into the Prudence folder if you chose that. It also reads message headers in your Sent folder to recognise the people you write to.
Prudence never sends email from your account, never deletes, never moves anything to spam or trash, and every move can be undone with "Put it back".
What we keep
- Verdicts, not messages. For each checked message we keep its Gmail id, the sender's website address, the verdict and the reasons shown to you, for 90 days. We do not store the text of your emails.
- Your personal context: the companies that write to you and their real addresses, recent orders and their carriers, your subscriptions, the people you write to, and data breaches you told us about. You can see and remove every item on the "What Prudence knows about you" page.
- What you share with a trusted person: only when you tap "Ask my trusted person", we keep the sender, subject, verdict and — only if you tick the box — the message text, so they can answer. Kept 90 days.
- Your Google tokens, encrypted, so we can keep checking new mail.
- A sign-in hint in your browser. A signed-in browser keeps a small cookie,
prudence_in, that says only that someone is signed in, so the welcome page does not flash before your screens; it is cleared when you sign out.
An early-access request. When you ask us to protect an address from our home page, we keep that address, whose it is (your parent's or yours), the first name and note if you gave them, and the language of the page, so we can add you as a tester by hand. The request also keeps which of our ads you came from, if any, and nothing else from the ad (no Facebook identifier). We keep it until you are added and for 90 days after, then delete it. Prudence sends no email about it: a person on our team writes to you.
Une demande d'accès anticipé garde aussi de laquelle de nos publicités vous venez, s'il y en a une, et rien d'autre de la publicité (aucun identifiant Facebook).
Who else sees it
- Meta Model API (Muse Spark): the message being checked and the relevant part of your context are sent to Meta to write the explanation, in Standard mode only — Meta does not use it to train its models. We never use Contributor mode.
- Domain registries (RDAP): we look up how old a link's website address is. Only the address is sent.
- abuse.ch URLhaus: we ask whether a link's website is on its public list of malware and phishing sites. Only the website's name is sent, never the full link or anything about you.
- Your trusted person: only what you choose to share, one message at a time.
- Nobody else. We do not sell, rent or use your data for advertising.
Google API Services User Data Policy
Prudence's use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. Gmail data is used only to provide the scam checks you see, never to train general-purpose AI models, and never transferred except as described above.
Deleting everything
Settings → "Delete my account" stops all checks, revokes Prudence's access to your Google account, and erases everything we stored. Your emails stay where they are. You can also remove access at myaccount.google.com/permissions.