TRUSTLAYER AI is a deterministic security layer and human authorization gateway designed to protect autonomous AI commerce agents from executing unauthorized transactions, falling victim to prompt injection attacks, or accepting hidden fees. Built with a React 18, Vite, and Tailwind CSS frontend alongside a FastAPI Python backend, the system parses natural language shopping requests into strict technical constraints, independently scans catalog listings using deterministic policy rules to generate 0–100 Trust Scores, and hardware-locks checkout on unsafe products. To complete a purchase, it enforces a mandatory human authorization gate that issues a single-use cryptographic token, enabling fail-closed USD PayPal Sandbox payments while preventing replay attacks and logging all activity to a tamper-evident audit ledger. Overcoming key integration challenges like CORS handling, popup blockers, and strict currency alignment, the project demonstrates how to safely bridge generative AI with secure payment infrastructure—with future plans to expand into multi-gateway support, browser extension SDKs, and semantic vector threat detection.

Built With

Share this project:

Updates

Submission history