Proposal: Encrypted Persistent AI Context
Author: Luther Lee Pollok III (@gamilu)
Summary
A standard for user-owned, encrypted AI context on Hedera. An individual's AI conversation history, preferences, and accumulated knowledge are encrypted with a key derived from an HTS token they own and a wallet signature they produce. The encrypted context is stored via HCS-1. No server ever holds plaintext. The key is never stored — it is derived fresh each session.
Problem
Every major AI platform today maintains custody of user context. Users cannot own their AI memory, cannot port it across providers, and are subject to platform compellability (government subpoenas, terms changes, shutdowns).
Solution
- Ownership layer: HTS NFT — token transfer = vault transfer
- Key derivation: HKDF-SHA256(token_id || wallet_signature, info="anamnesis-v1")
- Encryption: AES-256-GCM with token_id as AAD
- Storage: HCS-1 (recommended over HFS for per-session update cost)
- Audit trail: HCS topic
Relationship to Existing Standards
Builds on HCS-1 (storage), HTS (ownership), and complements HCS-11 (AI agent profiles) and HCS-14 (AI agent communication).
Reference Implementation
Full specification and Apache 2.0 reference implementation:
https://github.com/anamnesis-protocol/anamnesis
Draft PR
Companion PR forthcoming with full spec in docs/standards/.
Proposal: Encrypted Persistent AI Context
Author: Luther Lee Pollok III (@gamilu)
Summary
A standard for user-owned, encrypted AI context on Hedera. An individual's AI conversation history, preferences, and accumulated knowledge are encrypted with a key derived from an HTS token they own and a wallet signature they produce. The encrypted context is stored via HCS-1. No server ever holds plaintext. The key is never stored — it is derived fresh each session.
Problem
Every major AI platform today maintains custody of user context. Users cannot own their AI memory, cannot port it across providers, and are subject to platform compellability (government subpoenas, terms changes, shutdowns).
Solution
Relationship to Existing Standards
Builds on HCS-1 (storage), HTS (ownership), and complements HCS-11 (AI agent profiles) and HCS-14 (AI agent communication).
Reference Implementation
Full specification and Apache 2.0 reference implementation:
https://github.com/anamnesis-protocol/anamnesis
Draft PR
Companion PR forthcoming with full spec in
docs/standards/.