PII Redactor

No. Nothing is uploaded. The AI name detector runs entirely in your browser via WebAssembly, and the email, phone, card, SSN, IP and URL detection is plain pattern matching that also runs locally. Your text never leaves your device. Only the AI model is downloaded, once, then cached.

Two layers. An AI model (bert-base-NER) detects people, organizations and locations from context. Regular expressions detect emails, phone numbers, credit-card numbers, US Social Security numbers, IP addresses and URLs. You can turn each type on or off before redacting.

Block replaces each item with a run of solid blocks so the length is hidden. Label swaps in a readable tag like [EMAIL] or [PERSON] so a reader knows what was removed. Pseudonym assigns stable placeholders such as Person 1 and Email 2, and the same original value always reuses the same placeholder, which keeps anonymized text readable.

They are deliberately conservative to avoid false positives. Credit-card candidates must pass the Luhn checksum that real card numbers use, and SSNs must use a valid structure (for example, never area 000, 666 or 900 and above). This means a stray number in your text is unlikely to be flagged, but always review the output before sharing.

No automated redactor is perfect. The AI model misses unusual names and may mislabel rare words, and the patterns target common formats rather than every regional variant. Treat this as a fast first pass that removes the obvious personal data, then review the redacted text yourself before relying on it.