Prompt Guardian

01 · Problem

Employees paste API keys, credentials and confidential business content into public AI tools like ChatGPT and Claude, often by accident. One careless prompt can expose secrets, and companies are left choosing between banning AI tools and risking a leak.

02 · Solution

Prompt Guardian is an enterprise AI-security Chrome extension. It checks what a user is about to send and detects and blocks sensitive data, such as API keys, credentials and confidential content, before it leaves the browser. Admins manage device enrollment from a dashboard and get audit logs of every incident, so companies can keep using AI tools without exposing their secrets.

03 · Process

1. 01

   Research

   Studied how sensitive data leaks into AI tools and where existing safeguards fall short.

2. 02

   Project requirements

   Defined what must be detected and blocked, how admins enroll devices and what the audit trail must record.

3. 03

   System design

   Designed the system architecture, the ERD and the data model for devices, rules and incident logs.

4. 04

   API

   Built the FastAPI backend with detection logic for secrets, device enrollment and audit logging.

5. 05

   User interface

   Built the Chrome extension that checks prompts, and the admin dashboard with login, rules and audit pages.

6. 06

   Testing

   Tested detection against real secret patterns and worked on making it stronger.

7. 07

   Deployment

   Deployed the backend API on Render. The extension runs in the browser and is not hosted.

04 · Tech stack used

• FastAPI
• Python
• Next.js
• Chrome extension
• Secret detection (regex + entropy)
• Audit logging
• Render

05 · Links