Loading…
PrivacyAsst is an open-source framework that safeguards user privacy when tool-using large language model (LLM) agents access external tools and data. It uses privacy-preserving techniques (including TenSEAL) to encrypt or secure data during LLM-tool interactions, preventing leakage. The repository is the official implementation of the IEEE TDSC 2024 paper "PrivacyAsst: Safeguarding User Privacy in Tool-Using Large Language Model Agents."
AI named PrivacyAsst in February 2026.