Privatiser – Redact secrets, IPs, and PII before sharing with AI

Name: Privatiser – Redact secrets, IPs, and PII before sharing with AI
Availability: InStock
Author: r0otie

by r0otie·Feb 17, 2026·1 point·3 comments

Visit Project View on HN

AI Analysis

●●SolidSolve My ProblemSlick

The Take

Consistent, reversible pseudonymization with a downloadable mapping is the feature that actually makes this usable — you can paste anonymized text into ChatGPT and restore everything with one click. Shipping a browser extension that auto-anonymizes/pastes and a CLI/Python library for pipelines shows the author thought about real workflows. It relies on pattern-based detection and will miss exotic formats, but the local-only model and restore UX are a practical win.

Post Description

I built Privatiser, to allow the pasting of configs, logs, and code into ChatGPT/Claude but without risk leaking API keys, IPs, passwords, and customer data.

It detects and replaces sensitive values with consistent pseudonyms — so the AI can still reason about your text (e.g., "these two servers share the same IP") without seeing the real data. Then you paste the AI's response back and restore everything with one click.

What it catches: API keys (AWS, OpenAI, GitHub, Slack, etc.), JWTs, bearer tokens, SSH keys, connection strings, IPs, domains, emails, phone numbers, credit cards, SSNs, AWS ARNs/account IDs, Azure/GCP identifiers, UUIDs, and 200+ keyword-based patterns for passwords, hostnames, usernames, database names, etc.

It also understands natural language context — "password is secret123" and "host set to prod-db.internal" both get caught, not just key = "value" syntax.

Everything runs locally in your browser. No servers, no telemetry. Available as a Chrome/Firefox (Firefox is currently pending review) extension and a web tool.

Would love feedback on patterns you'd want added or things it misses.