PromptWard
Your prompts leak PII. This catches it before it leaves the browser.
A Chrome extension that intercepts sends to ChatGPT, Claude, Gemini, and Perplexity, runs on-device detection for names, SSNs, card numbers, emails, and phones — and shows a side-by-side redaction review. No server, no telemetry.
A local ONNX token-classification model (Rampart, from National Design Studio) runs entirely inside the extension — layered with deterministic heuristics for SSNs, card numbers, emails, and phone numbers — to find PII with no network call and no server round trip.
- 100% local detection and redaction - no prompt text ever leaves the device.
- Censor-by-default review flow with a 5s auto-confirm and explicit opt-out.
- Reversible placeholders: redacted tokens rehydrate back to the original values within the same conversation.
- Works across rich-text composers (Lexical/ProseMirror-style editors), not just plain textareas.
Intercept the send
Content scripts hook the composer on ChatGPT, Claude, Gemini, Perplexity, and Mistral — textareas, contenteditable, and Lexical/ProseMirror-style rich-text editors alike.
Detect on-device
A local ONNX token-classification model plus deterministic heuristics (Luhn checks, SSN and phone patterns) find PII with no network call and no telemetry.
Review, then send
A side-by-side diff shows original vs. redacted. The safe version auto-sends after 5 idle seconds; "Send original" is an explicit opt-out, never silent. Fail-closed if redaction can't be confirmed.