Oasis Foundation Awards Grant to Zeph for Combining TEEs and AI Companions
Zeph is enabling truly private conversations with AI models through confidential computing on the Oasis Network.
This Oasis grant recipient pushes beyond API wrappers to enable truly private conversations with AI models through confidential compute on the Oasis Network.
- Combining Nvidia confidential compute with TEEs for truly private AI interactions
- Zeph deploys local LLMs secured by Sapphire's confidential computing capabilities
- Personal companion powered by local AI and hardware-level privacy
Most AI companions are powered by APIs and send your private conversations to remote servers. Zeph is changing that. The Oasis Foundation has awarded a grant to support Zeph's vision of running AI models locally within Sapphire's secure computing environment. By deploying local language models within Oasis’ Trusted Execution Environments, Zeph is building what many thought impossible: an AI companion that keeps conversations verifiably private.
Zeph will use the grant to develop infrastructure that fundamentally reimagines how AI companions operate in Web3. Unlike traditional chatbots that rely on centralized AI providers, Zeph deploys fine-tuned open-source models locally, leveraging Sapphire's security through Trusted Execution Environments (TEEs) and Nvidia’s Confidential Compute support for H100s/H200s. This approach addresses a major issue in today's AI landscape: personal conversations flow through third-party services, exposing user data to potential breaches.
Notably, the AI companion app Muah.ai experienced a data breach affecting 1.9 million users, highlighting the vulnerabilities inherent in such platforms. By leveraging confidential computing on Sapphire, Zeph creates a secure environment where users can engage with AI companions without fear of their personal data being shared or sold.
Zeph's journey began at ETHGlobal Singapore, where the three founders saw an opportunity to make AI interactions truly private with Oasis. The team combines Kunal's cybersecurity expertise from his time securing Web3 protocols, Dylan's experience building scalable systems across Web2 and Web3, and Absinthe's award-winning product design background - including four Web3 hackathon wins and work with global brands like Binance. Together, they're tackling one of tech's biggest challenges: making private AI interactions not just possible, but accessible to everyone.
Learn More About Zeph
