Prompt PII Redactor

Strip personally-identifiable information from text before sending to an LLM. Pattern-based detection of emails, phone numbers, US SSN, credit cards, IP addresses, and (heuristic) names. Browser-only — your sensitive data never leaves the page.

How to use the Prompt PII Redactor

Paste text. Toggle which categories to detect. Replacement mode: placeholders preserve type ([EMAIL]); hash preserves identity (same value redacts to same hash — useful if you need to correlate without exposing the raw value); REDACTED is the most opaque.

Pattern-based PII detection is limited — it catches most well-formatted data but misses creative variants. For high-stakes redaction, use Microsoft Presidio or AWS Comprehend Medical instead.

Stripping PII before it reaches an LLM

Text headed to a third-party LLM API often carries personal data you'd rather not send: email addresses, phone numbers, US Social Security numbers, credit card numbers, IP addresses. This redactor detects those with pattern matching and replaces them before the text leaves the page — entirely in your browser, so the sensitive original is never transmitted.

You choose how each match is replaced: typed placeholders like [EMAIL] that preserve what kind of data was there, a deterministic SHA-256 hash that lets you correlate repeats without exposing the value, or an opaque [REDACTED]. Pattern matching catches well-formatted data but misses creative variants, so for high-stakes work pair it with a dedicated service. To replace named people and companies instead, use the prompt anonymizer.

Common use cases

• Pre-API scrubbing — remove emails, phones, and card numbers before sending text to a model.
• Log sanitizing — strip PII from logs or support tickets pasted into a prompt.
• Deterministic hashing — mask values while still matching repeats of the same one.
• Type-preserving review — keep [EMAIL] and [PHONE] labels so the text stays readable.
• Privacy by default — redact locally so raw data never leaves the browser.