End-to-end encryption for AI
Zama is an Open Source framework for securing AI applications in the
cloud.

Train your model anyway you want.
Zama secures the inference.
Zama secures the inference.
Using breakthrough homomorphic encryption, Zama enables any trained network, regardless of its architecture or training method, to run inference on encrypted user data.

Integrating Zama is a
seamless process
seamless process
Protecting your users’ Privacy should not require expertize in cryptography.
Zama automates the conversion and deployment of your trained neural network into a secure, cryptographic equivalent.
Compile
Compile your trained neural network into a homomorphic program

Deploy
Deploy a homomorphic runtime to execute your compiled program

Query
Integrate the client side SDK to securely query the homomorphic runtime

Features

Hardware accelerated

Microservice oriented

No retraining necessary

Open Source

Works with any neural network

Written in Rust
Create your own homomorphic
programs with
Concrete,
the high performance FHE library
programs with
Concrete,
the high performance FHE library
Concrete is an easy to use FHE library for cryptographers and engineers.
It enables fast prototyping and developement of complex homomorphic
applications, such as Zama’s AI framework.



Programmable Bootstrapping
Evaluate any function homomorphically, in only a few milliseconds.

Written in Rust
Concrete is written in rust, making fast, reliable and easily
extensible.

Open Source
Concrete is available in open source. We welcome contributions of
any kind!