Humane Proxy

Humane Proxy is a lightweight, plug-and-play AI safety middleware designed to protect humans when interacting with Language Learning Models (LLMs). Acting as a reverse proxy between users and LLMs, it intercepts messages to detect self-harm ideation or criminal intent, blocks harmful content, sends empathetic responses, and alerts human operators through preferred channels. With a multi-stage cascade (heuristics, semantic embeddings, and reasoning LLM), it offers high recall and low false positives. Designed for developers, companies, and anyone integrating LLMs, it ensures privacy and compliance by never storing raw message content and by supporting audit logs and erasure.