GuardianAI scans everything you type into Claude and ChatGPT in real time — looking for sensitive data before it leaves your machine. ID numbers, passwords, API keys, email addresses. If it finds something, it warns you before you share it accidentally.
🇿🇦
Enable POPIA mode
Tags detections with POPIA sections (§11, §19, §26) and applies stricter defaults. Recommended for South African users.
🔒Everything runs locally in your browser. No data is sent anywhere. No cloud, no login, no storage of your prompts — ever. Privacy Policy ↗
Three levels. Always your call.
GuardianAI uses three risk levels depending on what it finds.
HIGH
Submission blocked
Credentials, ID numbers, API keys, private keys. A red banner appears — you can't send until you dismiss or replace the sensitive content.
MEDIUM
Amber banner
Email addresses, phone numbers, custom keywords. A modal shows exactly what was detected. Replace sensitive content in place, send anyway, or ignore and scan again later.
LOW
Silent
Source code patterns. Flagged internally — no interruption. You can send normally.
Let's make sure it's working.
Copy any value below and paste it into Claude or ChatGPT. You should see the matching warning within 1 second.
📋Test values — all fake, safe to use
8001015009087HIGH
AKIAIOSFODNN7EXAMPLEHIGH
test@company.comMEDIUM
Help me write a cover letterNone
⚙️Customise thresholds, add keywords, and configure auto-sanitize anytime from the GuardianAI popup in your Chrome toolbar.
✓You're all set. GuardianAI is now protecting your prompts.