EXO Labs Launches Privacy-Preserving Search for Local LLMs Using Homomorphic Encryption, Reducing Data Transfer by 100,000x

Jan 3, 2025, 12:07 PM

EXO Labs has introduced a new privacy-preserving web search feature designed for local large language models (LLMs). This innovation utilizes Linearly Homomorphic Encryption to allow local LLMs to access real-time data, such as weather updates and stock prices, without compromising user privacy. The system is reported to require 100,000 times less data transfer compared to traditional client synchronization methods, with a round-trip time of less than two seconds. The EXO Private Search enables users to run LLMs locally while simultaneously fetching updates from encrypted server queries, thus maintaining end-to-end privacy. This development addresses the growing demand for private search capabilities in the evolving landscape of AI and LLMs.