Category: confidential AI / FHE compute network / proof-of-human identity and agent infrastructure
Summary: Privasea is best cataloged as a multi-surface privacy-compute project that combines an FHE-oriented AI network, a proof-of-human / human-liveness product, node-operator infrastructure, and an emerging agent ecosystem. Its current official materials do not read like a single protocol-only stack: the website jointly markets the Privasea AI Network, the FheID/ImHuman app, DeepSea AI agents, and WorkHeart/Privanetix node participation. The through-line across those surfaces is a claim that fully homomorphic encryption and decentralized compute can support privacy-preserving AI inference while verified real humans, not bots, become a core trust input for agentic and social systems.
What it does:
Markets the Privasea AI Network as a decentralized privacy-compute network for encrypted machine-learning inference using FHE plus blockchain-based incentives
Offers the FheID / ImHuman app as a proof-of-human-liveness product that claims to verify human likeness without disclosing raw identity or storing usable biometric data on Privasea servers
Runs node-operator surfaces around Privanetix and WorkHeart nodes, showing that some of the project’s most concrete public docs are operational guides for contributors supplying compute and participating in the network
Positions DeepSea as an AI-agent ecosystem powered by verified human users, with early examples around social mining, staking, and data-oriented agent roles
Maintains a public GitHub org with an older Privasea-General overview plus HESEA and FHE-related repos that help confirm the project’s technical lineage beyond the marketing site
Key claims:
The AI Network pages say Privasea blends fully homomorphic encryption with blockchain incentives to enable confidential ML inference and privacy-preserving data collaboration
The network materials describe a hybrid Proof-of-Work and Proof-of-Stake model regulating gas-fee and staking mechanics for miners / operators
The FheID materials claim facial-biometric data is transformed into encrypted vectors on-device, with the project unable to access or decrypt the original data
The product surface now ties proof-of-human verification directly into downstream agent and reward systems, suggesting Privasea is trying to connect privacy-preserving identity with decentralized AI participation
The public GitHub repos show an open-source FHE library and demo lineage, which is useful context even though the most current node and product materials live on the website and docs
Whitepaper: An official Privasea AI Network whitepaper is linked from the site (/white-paper) and currently points to a Google Drive PDF. The most useful primary-source corpus in practice was the whitepaper landing page, the AI Network and FheID pages, node docs, and the public GitHub org; see ../whitepapers/privasea-primary-sources-2026-04-27.md.