PreSend
Local-first sensitive-data protection for AI chats.
PreSend detects credentials and semantically sensitive information before submission and warns the user before potentially sensitive content is sent.
Inspect locally, then let the user decide.
PreSend adds a review step before content leaves a supported AI chat composer. It warns rather than claiming perfect protection.
- Detect credentials with local deterministic checks
- Detect semantically sensitive information with local inference
- Warn before submission
- Offer Redact or Replace with Safe Example
- Keep feedback decisions and allowlisting local
No scanning backend.
Detection and semantic inference run in the browser. PreSend has no backend for scanning, telemetry, or feedback, and it does not use cloud inference. Detection can produce false positives or miss sensitive information, so the user remains in control of the final submission.
Microsoft Edge and Firefox.
PreSend is being prepared for Microsoft Edge Add-ons and Firefox Add-ons. Store links will appear here only after the listings are published.