Mithril Security

Mithril Security, founded in Paris in April 2021, operates in the confidential computing sector, offering solutions for software vendors to deploy AI in security-conscious organizations. The company was established by CEO Daniel Huynh, COO Raphaël Millet, and CTO Mehdi Bessaa. Huynh's background includes work at Microsoft on Privacy Enhancing Technologies, which inspired the company's focus on making AI more trustworthy. Millet's experience as a consultant for banks adopting AI highlighted the challenge of evaluating SaaS solutions without security compromises, a problem Mithril's technology aims to solve. The founding team was united by a goal to democratize privacy-preserving AI.

The company's core business revolves around enabling the use of AI with sensitive data by leveraging confidential computing and secure enclaves. This technology creates isolated, hardware-enforced environments where data can be processed in its decrypted state without being exposed to the underlying infrastructure, not even to system administrators. This approach provides organizations with security and control levels comparable to on-premise deployments while using cloud-based AI SaaS products. The business model targets AI providers and companies in regulated industries like healthcare, finance, and legal, which handle sensitive information and face stringent data protection regulations.

Mithril Security has developed several open-source tools, including BlindAI, BlindBox, and BlindChat. BlindAI is a solution for deploying AI models, such as those for analyzing medical images or documents, with privacy guarantees. It consists of a server that runs models within secure enclaves and a client SDK for users to query them securely. BlindBox was developed to protect data privacy when using Large Language Models (LLMs), ensuring user data remains confidential from any intermediary, including the model owner. BlindChat is a private AI assistant designed to provide end-to-end encryption for user interactions with conversational AI. In May 2023, the company announced it had raised €1.2 million in a pre-seed round led by Auriga Cyber Ventures, with participation from Polytechnique Ventures and IT-Farm. However, in the second half of 2024, Mithril Security was acquired by the French AI startup H Company. Following the acquisition, founder Daniel Huynh and part of the team joined H Company, though Huynh has since departed.

Keywords: confidential computing, AI security, secure enclaves, data privacy, privacy enhancing technologies, AI deployment, large language models, LLM security, zero-trust AI, open-source security, AICert, BlindAI, BlindBox, BlindChat, AI governance, encrypted data processing, machine learning security, trusted execution environments, secure AI SaaS, data protection