Vitalik Buterin's AI 'Paranoid Mode' Uses Tor to Mask Prompts
The Ethereum co-founder's privacy setup combines Tor, zkAPI and local prompt rewriting, though he says it slows new chats and still leaves gaps.
Ethereum co-founder Vitalik Buterin has assembled a privacy-focused configuration for using AI chatbots that he describes as a "paranoid mode," according to a BeInCrypto report. The setup is aimed at limiting how much an AI provider can learn about a user's prompts.
The configuration combines Tor for network-level anonymity, a component referred to as zkAPI, and local rewriting of prompts before they are submitted. Together the layers are meant to obscure both the content of a query and the identity of the person sending it.
Trade-offs acknowledged
Buterin said the approach slows down the process of starting new chats and that gaps remain in the protection it provides, per the report. The account did not detail how zkAPI is implemented or which AI services the setup is compatible with.
The experiment reflects wider interest in privacy for AI tools, where prompts can expose personal or business information to model providers. The report described it as a personal project and did not indicate a public release or packaged tooling.