We’re releasing HEVEC, a vector database built on homomorphic encryption, enabling end-to-end privacy with real-time search at scale.
HEVEC is designed as a drop-in alternative to plaintext vector databases and supports real-time encrypted search at scale (1M vectors in ~187 ms).
Key points: - A secure, drop-in alternative to plaintext vector databases - End-to-end homomorphic encryption for both data and queries - Real-time encrypted search at scale (1M vectors in 187 ms)
As personal AI agents become deeply personalized, data ownership must belong to users.
HEVEC enforces this through privacy-by-design architecture.
We’d appreciate feedback from the AI, systems, and privacy communities.
ddtaylor•1h ago
Is this closer to Fully Homomorphic Encryption (FHE) or partial?
cloneisme•1h ago
HEVEC uses partial homomorphic encryption, not FHE.
It supports only the operations needed for an encrypted vector database (search, insert, delete, etc.), which keeps performance practical.
Implementation details are in our paper: https://arxiv.org/abs/2506.17336
Happy to elaborate if helpful.