Mithril Security

Founded by Daniel Huynh in 2021

Protect sensitive AI training data and model intellectual property using secure hardware enclaves

Pricing model

Demo

Rating

★ People love it

Time to value

Moderate Setup (1-3 hours)

You can use Mithril Security to deploy and train artificial intelligence models while keeping datasets and model weights private. The system runs workloads inside secure hardware enclaves, preventing unauthorized parties or cloud providers from viewing confidential data. With its AICert technology, you can generate cryptographic proofs to trace model training origins and verify data integrity. It also allows developers to host large language models privately using hardened environments, ensuring user prompts and intellectual property remain protected from leaks or tampering across your infrastructure.

What Mithril Security does

Verify the training provenance of machine learning modelsDeploy private instances of open-source language modelsCreate cryptographic certificates for AI training proceduresIsolate sensitive inference workloads inside secure hardware enclavesPrevent unauthorized data visibility during cloud computationsAudit model code and infrastructure through open source repositoriesIsolates AI inference and training workloads inside hardware-backed secure enclavesGenerates cryptographic proofs to verify training datasets and algorithmsPrevents cloud providers and unauthorized users from reading private user promptsCreates unforgeable hardware identity records for machine learning models using TPMsDeploys open-source large language models with hardened end-to-end encryptionProvides audited open-source codebases for public security verification