Cowrie Research

privasea-primary-sources-2026-04-27

4 min read

Privasea primary-source snapshot

Last reviewed: 2026-04-27 UTC

Project status snapshot

What the official materials say

1) Homepage / overall product framing

  • URL: https://www.privasea.ai/
  • Key takeaways:
    • Privasea no longer presents as only an FHE research effort or only a blockchain network.
    • The homepage combines four major surfaces: the Privasea AI Network, the FheID/ImHuman proof-of-human app, DeepSea AI agents, and node-operator infrastructure.
    • This is the key clue that the project should be cataloged as a multi-surface privacy/AI/identity stack rather than a simple “AI chain.”

2) AI network page

  • URL: https://www.privasea.ai/ai-network
  • Key takeaways:
    • Privasea describes the network as a DePIN-style infrastructure layer for privacy-preserving AI.
    • The page explicitly says the network combines fully homomorphic encryption with blockchain incentives.
    • The network claims include confidential ML inference, data-value sharing instead of raw-data sharing, and a hybrid PoW/PoS incentive model.
    • The page also references Zama, which is useful context for understanding the project’s FHE orientation.

3) AI agent page / DeepSea

  • URL: https://www.privasea.ai/ai-agent
  • Key takeaways:
    • Privasea is tying its AI network to a separate agent ecosystem called DeepSea.
    • The site says the agents are powered by verified human users from the FheID/ImHuman app.
    • This is strategically important: Privasea is positioning “verified humans + private AI infrastructure” as a combined system, not isolated products.

4) FheID privacy page

  • URL: https://www.privasea.ai/fheid-app-privacy
  • Key takeaways:
    • The FheID materials say users verify human likeness through facial biometrics while Privasea avoids storing the raw biometric data on its own servers.
    • The page says encrypted vectors are generated locally and that Privasea cannot decrypt them.
    • Whether those claims hold in implementation is a separate question, but they are central to how Privasea wants the product categorized: privacy-preserving proof of human rather than ordinary KYC or identity collection.

5) Whitepaper landing page

  • URL: https://www.privasea.ai/white-paper
  • Key takeaways:
    • The site explicitly says a Privasea AI Network whitepaper exists and links to a downloadable document.
    • The landing page was more informative about the existence and currency of the whitepaper than about its detailed contents.
    • For this pass, the whitepaper page is best treated as proof that an official paper exists, while the website and docs provide the clearer operational/product narrative.

6) Docs hub / node operations

  • URL: https://docs.privasea.ai/
  • Key takeaways:
    • The current public docs are dominated by node-operation guides rather than protocol theory or SDK docs.
    • The docs emphasize Privanetix and WorkHeart node setup and operation across Windows, Ubuntu, and macOS.
    • This matters because it shows where the project’s most concrete public operating surface currently lives: contributor/node docs, not developer API documentation.

7) GitHub org / technical lineage

  • URL: https://github.com/Privasea
  • Key takeaways:
    • The public GitHub org still exposes the project’s cryptography lineage, including HESEA_Lib and older demo repositories.
    • Privasea-General describes the project as a privacy-computing solution using AI models and decentralized computation crowdsourcing.
    • HESEA_Lib is especially useful because it makes the FHE orientation concrete and not purely marketing-driven.
    • Some repos are forks or older experiments, so the org should be read as lineage/context rather than a perfect map of the live product stack.

Whitepaper / technical-doc status

  • An official Privasea AI Network whitepaper is linked from the site and currently resolves through Google Drive.
  • The most operationally informative sources in this pass were the AI Network page, FheID privacy materials, AI-agent page, node docs, and the public GitHub org.
  • The docs posture suggests that current public detail is strongest around node participation and product framing, not around an open developer API surface.

Research notes

  • Best current category for the library: confidential AI / FHE compute network / proof-of-human identity and agent infrastructure.
  • The important cataloging move is to avoid reducing Privasea to only one of its surfaces: it is simultaneously selling private AI compute, human-verification infrastructure, and an agent ecosystem.
  • Projects with this mix can look noisy from the homepage, but the combined site + privacy pages + node docs + GitHub lineage make the architecture and target audience much clearer.

Graph View