FinePrint.ai

FinePrint.ai

Inspiration

We all lie. We scroll to the bottom of the "Terms & Conditions" and click I Agree without reading a single word. But recently, I almost signed a rental lease that gave my landlord ownership of any code I wrote in my apartment. That terrified me. I realized that legal counsel costs $300/hour, but getting scammed costs even more. I wanted to build a "Legal Shield" for the 99%—an AI that reads the fine print so you don't have to.

What it does

FinePrint.ai is an intelligent contract auditor that protects users from predatory terms. It features three layers of defense:

** Risk Scanner:** Instantly audits PDFs (Leases, NDAs, Employment Contracts) and assigns a "Safety Score" (0-100). It highlights specific "High Risk" clauses like hidden fees, IP theft, or invasive privacy violations.
** Roast Mode:** A sarcastic, viral-ready feature where the AI acts as a comedian to mock the most ridiculous terms in the contract.
** The Negotiator:** Turns anger into action by auto-drafting a professional legal email to the counterparty, requesting the removal of unfair terms.

⚙️ How we built it

The app is built in Python using Streamlit for the frontend interface. The core intelligence is powered by Google Gemini 1.5 Flash (via the google-genai SDK). We chose Gemini 1.5 Flash for its massive context window—allowing it to read entire leases in one pass—and its low latency. We used PyPDF2 for text extraction and Streamlit Session State to ensure a smooth, non-reloading user experience.

Challenges we ran into

The biggest challenge was "Context Awareness." A Non-Compete clause is standard in an Employment contract but predatory in a Lease. Initially, the AI flagged everything as "High Risk." We solved this by implementing a Dynamic Benchmarking System in our prompt engineering, which first detects the document type (e.g., "Residential Lease") and then compares clauses against market standards for that specific type.

Accomplishments that we're proud of

We are most proud of the Scoring Engine. Seeing the AI correctly assign a "0/100" to a predatory lease and a "95/100" to a fair one—without us hard-coding the rules—was a huge win. We are also proud of the "Roast Mode", which makes a boring legal tool feel fun and engaging.

What's next for FinePrint.ai

The next step is to build a Browser Extension that automatically scans "Terms of Service" before you click "Checkout" on e-commerce sites. We also plan to integrate with DocuSign to flag risks directly inside the e-signature workflow.