PII Sanitizer & Data Redactor, LLM Prompt Anonymizer & Token De-Anonymizer
Sharing customer support tickets, error logs, and proprietary communications with Large Language Models (ChatGPT, Claude, Gemini) introduces significant privacy risks and GDPR/HIPAA compliance violations if Personally Identifiable Information (PII) is exposed. The PII Sanitizer redacts sensitive identifiers (emails, phone numbers, IP addresses, SSNs, credit card numbers) using reversible surrogate tokens, allowing users to safely query AI models and restore original identities in AI responses.
A customer support specialist prepares an error ticket for an AI summarization prompt: Customer John Doe (email: [email protected], phone: 555-019-2834, IP: 192.168.1.45) reports payment failed with card 4532-0150-1234-5678.. Clicking Sanitize & Redact PII detects the PII entities, validates the Visa card number using the Luhn algorithm, and replaces them with surrogate tokens: Customer John Doe (email: [EMAIL_1], phone: [PHONE_1], IP: [IP_1]) reports payment failed with card [CREDIT_CARD_1].. An in-memory mapping registry records the surrogate relationships. The specialist pastes the sanitized text into ChatGPT. When the AI returns a summary referencing [EMAIL_1], pasting the response into the De-Anonymize / Restore panel instantly swaps original identifiers back into place.
Entity detection, Luhn validation, and surrogate mapping run locally in browser memory, ensuring sensitive PII is never exposed to remote logging servers.
Core Architecture & Mathematical Formula
Sanitization: Text ➔ Regex Pattern Scan [Email, Phone, IP, SSN] ➔ Luhn Checksum [Credit Card] ➔ Token Substitution [TOKEN_N]
Scans text for PII patterns; validates credit card numbers via Luhn Mod-10 algorithm; establishes bidirectional surrogate token mapping in volatile browser memory.
Best Practices & Essential Guidelines
- Sanitize Prompts Before Submitting to Third-Party AI Services: Never paste raw customer medical data, banking numbers, or contact records into public LLM interfaces; surrogate tokenization preserves context while safeguarding identities.
- Verify Credit Card Validation via the Luhn Algorithm: The sanitizer uses Luhn checksum verification to prevent false-positive redaction of general 16-digit order or tracking numbers that are not genuine credit cards.
- Maintain the Reversible Surrogate Registry for Response Mapping: Keep your browser session active while chatting with AI models so you can paste the AI's generated response into the De-Anonymize panel to restore original customer names.
- Customize Redaction Categories Based on Compliance Scope: Toggle specific entity filters (such as disabling IP address redaction if debugging network routing) to tailor the sanitization scope to your specific workflow.
