LLMTor – Privacy Layer for Anonymous Access to Frontier LLMs

Summary: LLMTor provides a cryptographic privacy layer that separates user identity from prompts sent to public large language models (LLMs). It uses blind RSA signatures and Tor to ensure neither the LLM providers nor LLMTor operators can link prompts to users, enabling anonymous access to models like ChatGPT.

What it does

LLMTor acts as a proxy between users and LLM providers, issuing blind RSA-signed tokens that users redeem anonymously over Tor. This decouples user identity from prompt content, preventing linkage by either the proxy or the LLM servers.

Who it's for

It is designed for users who want to access public frontier LLMs without revealing their identity or prompt data to providers or intermediaries.

Why it matters

LLMTor addresses privacy concerns by preventing LLM providers and operators from associating prompts with individual users, unlike direct access where plaintext prompts are exposed.