Privacy Scrubber – Zero-Trust PII Sanitizer

Enterprise teams in insurance, finance, and legal risked massive compliance fines by pasting confidential documents into ChatGPT, as Cloud DLP tools added secondary exposure points.

• 87% of enterprises lack technical controls to prevent PII from entering GenAI (K2view 2026 data), leaving companies vulnerable during the 471M breach wave.
• Server-side DLP proxies (like Integral Privacy or cloud gateways) create catastrophic supply-chain attack vectors, as proven by the LiteLLM breach leaking terabytes of secrets.
• Gartner projects that by 2029 majority of privacy incidents will stem from AI inferences rather than direct leaks, requiring mandatory client-side pre-processing.

Invented Zero-Trust Data Sanitization (ZTDS) to run all PII detection entirely in volatile RAM. Deployed a Chrome extension, offline PWA, and an MCP server for secure AI agent integration.

1. 1Invented Zero-Trust Data Sanitization (ZTDS): all PII detection and masking runs in volatile browser RAM. No data ever leaves the device. A mathematical guarantee, not a policy promise.
2. 2Built a 20+ entity detection engine: Names, Emails, Phones, SSNs, Credit Cards, IBANs, Medical Record Numbers, API Keys, Passwords, IPs with 17+ industry profiles (HIPAA, GDPR, SOC 2, Legal, Finance, HR, Developer).
3. 3Shipped Chrome and Firefox extensions working automatically across ChatGPT, Claude, Gemini, Copilot, Grok, DeepSeek, and Qwen, scrubbing in real-time before text reaches AI servers.
4. 4Published an MCP Server on npm: local AI agents call scrub_pii() and reveal_pii() as native tools. Zero data leaves the machine. Air-gapped compatible.
5. 5Reversible tokenisation: PII is replaced with [NAME_1], [CARD_2] tokens. The AI reasons with full context intact. One click restores originals from the local session map.