HN Showcase
Back to browse
PrivacyShield – Mask your PII before it reaches ChatGPT/Claude

PrivacyShield – Mask your PII before it reaches ChatGPT/Claude

by BuddhikaR·Mar 3, 2026·1 point·1 comment
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemDark HorseSlick

Local PII masking for ChatGPT—free and zero-server, but detection accuracy unverified.

Strengths
  • •Genuinely solves a painful, real problem: accidental PII leakage to third-party AI APIs.
  • •100% local processing with zero servers removes trust friction—verifiable in DevTools.
  • •Frictionless UX: one-click install, passive detection, review-before-send workflow respects user agency.
Weaknesses
  • •PII regex detection (15+ types claimed) is brittle—high false negatives on obfuscated formats, no ML-backed validation shown.
  • •v0.1 with acknowledged bugs; no benchmarks on detection rates or false positive impact on usability.
Category
Security
Target Audience

Individual AI users, freelancers, professionals handling sensitive data

Similar To

Ghostery · Privacy Badger · DuckDuckGo extension

Post Description

Hi HN, I built this because I kept pasting client data and API keys into ChatGPT without thinking.

It detects 15+ types of PII as you type, masks them with placeholders like [PERSON_A] before sending, then unmasks the AI's response. Everything runs locally — no servers, no data collection, zero network requests. You can verify in DevTools.

Built as a solo project using Claude Code.

Chrome Web Store: https://chromewebstore.google.com/detail/privacyshield/nklgh...

This is v0.1 — would love feedback. Bugs: [email protected] or GitHub issues.

Similar Projects

Security●●Solid

FunWithText – Free browser PII sanitizer and prompt injection scanner

Runs fully in the browser for regex-based redaction of structured PII (emails, cards, phones, API keys, IPs, SSNs) and offers an optional Claude step to catch names/addresses patterns miss. Clear placeholders like [NAME] and a privacy toggle make it practical for quick, manual workflows, and I like the Norwegian NIN/phone support. It's useful and thoughtfully privacy-first but not novel — the real win would be integrations (batch processing, editor/extension) or independent accuracy metrics for the AI step.

Solve My ProblemSlickNiche Gem
atbj
205mo ago