π‘οΈ Six to 7 Security
π§ Inspiration
In todayβs data-driven world, organizations handle massive amounts of documents, contracts, reports, and internal memos with little visibility into whatβs actually sensitive. Manual review is slow, inconsistent, and risky.
We wanted to build an AI-powered regulatory document classifier that could instantly detect, label, and secure sensitive information. Thatβs how Six7 Security was born, an intelligent assistant for document risk detection and compliance automation.
π What It Does
Six7 Security automatically analyzes uploaded files and classifies them into one of four sensitivity levels:
- π’ Public β safe, shareable content
- π‘ Confidential β internal or restricted business information
- π΄ Highly Sensitive β contains PII or proprietary material
- β οΈ Unsafe β non-compliant, exploitative, or harmful content
β¨ Key Features
- π Multi-format uploads (PDFs, text, and images)
- π€ AI classification powered by dynamic prompts
- π§ Dual verification using two LLMs for higher confidence
- π Citation-based evidence showing exact pages or images that triggered the classification
- π©βπΌ Human-in-the-loop review for oversight
- π Interactive dashboard for file management and visualization
βοΈ How We Built It
π» Frontend
- Built using React, Tailwind CSS, and Framer Motion
- Includes a landing page with a βGet Startedβ button that transitions to the upload dashboard
π§© Backend
- Implemented with Flask (Python) for API routing
- Used PyPDF2 to extract text from PDFs and images
- Integrated OpenAI GPT and Gemini for classification and reasoning
π‘ What We Learned
- How to combine LLMs, OCR, and reasoning for compliance and risk management
- The value of prompt modularity and confidence-based consensus
- The importance of explainable AI in sensitive content detection
- Why clear UX builds trust in AI-driven security systems
Six7 Security aims to redefine how organizations detect, classify, and protect sensitive data β automatically, intelligently, and securely. π

Log in or sign up for Devpost to join the conversation.