Inspiration

People use AI every day to understand legal documents, insurance papers, medical reports, and contracts. The problem is that these documents contain personal information like names, phone numbers, addresses, email IDs, and account numbers. Many people don't realize they are sharing this private information with AI tools. My solution help people use AI safely without exposing their personal information.

What it does

rivateDocs AI helps people understand sensitive documents while protecting their privacy. A user uploads a document, and the application automatically finds personal information like names, phone numbers, email addresses, and other sensitive details. The user can also manually hide anything the system misses. Only the cleaned version of the document is sent to AI for analysis. The original document and personal information never leave the user's control. The AI then explains or summarizes the document without seeing the user's private information.

How we built it

Frontend using React, TypeScript, and Tailwind CSS to create a simple and user-friendly interface. The backend is built with Node.js and Express. I used Google Gemini to analyze the cleaned document and designed the project so it can work with Midnight to securely store a privacy receipt without storing the actual document or personal information.

Challenges we ran into

One of the biggest challenges was making sure private information never reached the AI. Automatically finding sensitive information is not always enough because every document is different. To solve this, we added a manual review step where users can hide any additional information before sending the document to AI. Another challenge was making sure the AI only analyzed the cleaned document and never tried to guess the hidden information.

Accomplishments that we're proud of

I built a complete privacy-first workflow from start to finish. What my application do: Detects sensitive information automatically. Lets users manually hide additional information. Sends only the cleaned document to AI. Gives users useful document analysis while protecting their privacy. Creates a foundation for secure privacy verification using Midnight.

What we learned

I learned that protecting privacy is much more than hiding names or phone numbers. Users should always stay in control of their own data. Learned how privacy-focused technologies like Midnight can help build applications that are both useful and trustworthy.

What's next for PrivateDocs AI

Want to support more document types like Word files and scanned PDF and I want to improve the detection of sensitive information using smarter AI techniques.

Share this project:

Updates