PromptShield — Enterprise AI Security & DLP

Problem Statement

Teams paste customer data, credentials and internal documents into AI tools every day, and most organisations can't see which AI services are in use, let alone control what leaves the building.

Solution

• •An inspection gateway that checks outbound prompt payloads and redacts sensitive data — such as NUBAN, BVN and card numbers — before it reaches a model.
• •A Chrome (Manifest V3) extension that enforces the organisation's AI policy: unsanctioned AI tools are blocked and every attempt is reported back to PromptShield.
• •An admin console for policies, shadow-AI discovery, incidents, risk, compliance and approvals.

Technologies Used

• •Backend: Django, Django REST Framework, Simple JWT, django-ratelimit, cryptography
• •Console: Next.js, React, TanStack Query, Recharts, Tailwind CSS
• •Extension: Chrome Manifest V3 (declarativeNetRequest, service worker)
• •AI: Google Gemini API

Key Features

• •Prompt inspection and redaction of sensitive data
• •Blocking of unsanctioned AI tools via the browser extension
• •Shadow-AI discovery across the organisation
• •Policy engine with approval workflows
• •Incident, risk and compliance tracking
• •Company registration and enterprise sign-in