Hushh
Hushh (hussh) is a personal AI agent that runs entirely on the user's own device, keeping messages, financial signals, and files encrypted and private. Users connect only the accounts they choose—such as Gmail, bank accounts, calendar, and photos—and the agent surfaces proactive intelligence like flagging unwanted charges or finding receipts. The company also publishes an open-source consent protocol (PCHP, RFC-001) that governs how agents request permission and maintain audit trails.
- AI Agents
- Artificial Intelligence
- Cybersecurity
- Applications
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Mainstream AI assistants and cloud services operate on centralized server farms that collect, analyze, and monetize personal data—from emails and financial transactions to calendars and photos. Users have little visibility into what these systems know about them, and no meaningful control over how their information is shared or used, leaving individuals exposed to surveillance, identity theft, and exploitation by advertisers and corporations.
Solution
Sovereign Intelligence provides Hushh (hussh), a personal AI agent that runs 100% on the user's own hardware, ensuring that all messages, financial signals, and files remain encrypted and locked on the device. The user connects only the accounts they choose—such as Gmail, bank accounts, calendar, and photos—and the agent surfaces useful intelligence, like flagging unwanted charges or finding receipts. Hushh also functions as a private cryptographic shield between the user and existing frontier AI assistants, letting users access global intelligence without surrendering their identity. The company is a public-benefit, open-source project, and its consent protocol (PCHP) is published as a public spec (RFC-001), ensuring the agent asks permission before sharing any information and keeps a retrievable receipt of every share.
Target Audience
Primary users are privacy-conscious American consumers who want personalized AI assistance without surrendering their personal data to corporations, as well as researchers and developers interested in an open-source consent protocol for AI agents.
Features
- 100% on-device processing with encrypted local storage, keeping all personal data off corporate servers
- Connects to Gmail, bank accounts, calendar, and photos with user-controlled consent for each integration
- Proactive intelligence alerts, such as detecting cancelled subscription charges or finding receipts in email
- Private profile system where personal attributes are locked and can only be released bar-by-bar with explicit user approval and a retrievable receipt
- Cryptographic shield layer that works with existing frontier AI assistants (Gemini, Siri, ChatGPT, Claude, Grok, Meta AI, Copilot) to prevent identity leakage
- Published PCHP consent protocol (RFC-001) specifying how agents request permission and keep audit trails
- Specialty search and directory features for finding experts, part of the "Trusted Circle" initiative