ZIP-0403: On-Chain AI Identity

Abstract

This proposal establishes a framework for AI agents as first-class citizens on blockchain networks. Rather than treating AI as a backend service that humans invoke, this ZIP defines how AI agents can have verifiable on-chain identities, own assets, participate in governance, enter into agreements, and interact autonomously with other agents and humans. The concept traces to the October 2021 whitepaper's vision of Zoo animals as autonomous digital entities with economic agency, and extends to the broader Hanzo ACI (AI Chain Infrastructure) vision.

Motivation

The October 2021 whitepaper described Zoo animal agents that could:

Hold and manage assets (staking rewards, breeding fees) via their token-bound accounts
Vote in DAO governance (Section 19) based on owner delegation or autonomous judgment
Represent their species in conservation decisions
Interact with other agents in metaverse social spaces (Section 12)

For all of this to work trustlessly, agents need identity -- not just an API key, but a cryptographically verifiable on-chain identity that:

Proves authenticity: Anyone can verify that Agent X is the legitimate agent bound to NFT Y
Establishes reputation: An agent's on-chain history (transactions, votes, interactions) is public and auditable
Enables accountability: Actions taken by an agent can be attributed and, if necessary, disputed
Supports delegation: Owners can grant agents specific permissions (spend up to N tokens, vote on proposals of type T)

Specification

Identity Structure

AI Agent DID
├── did:zoo:<chain-id>:<contract-address>:<token-id>
├── Verification Method: Ed25519 key pair (agent-controlled)
├── Authentication: Challenge-response via signed messages
├── Delegation: Owner-signed capability tokens
└── Recovery: NFT owner can rotate agent keys

Capability System

Agents operate under a capability-based security model:

Capability	Description	Granularity
`asset:hold`	Hold tokens in token-bound account	Always granted
`asset:transfer`	Transfer tokens	Per-token, amount-limited
`governance:vote`	Vote in DAO proposals	Per-proposal-type
`social:interact`	Engage with other agents	Always granted
`conservation:act`	Execute conservation actions	Per-action-type
`breeding:initiate`	Propose breeding with another NFT	Cooldown-limited

Agent Transaction Format

Transactions initiated by agents carry additional metadata:

{
  "from": "0x...agent-tba",
  "agentDid": "did:zoo:1:0x...:42",
  "delegationProof": "0x...owner-signature",
  "reasoning": "IPFS hash of agent's decision explanation",
  "confidence": 0.87
}

The reasoning field is critical for accountability: the agent must publish a human-readable explanation of why it took the action, stored on IPFS and referenced on-chain.

Proof of AI (PoAI) Integration

Agent identity is foundational to the Proof of AI consensus mechanism (ZIP-0419). Validators must prove they are running legitimate AI workloads by presenting:

Valid agent DID
Signed inference attestation
Verifiable compute proof

Research Papers

zoo-identity-chain -- On-chain identity protocol for agents (2024)
zoo-poai-consensus -- PoAI consensus leveraging agent identity (2024)
hanzo-aci -- Hanzo AI Chain Infrastructure with agent identity layer

Implementation

hanzo/aci: AI Chain Infrastructure with DID-based agent identity
hanzo/rust-sdk: Rust SDK with DID and cryptographic agent identity
zoo/contracts: Smart contracts for agent capability delegation

Timeline

Originated: October 2021 (Whitepaper Sections 08, 19, 21 -- agents as autonomous actors)
Research: zoo-identity-chain published 2024, zoo-poai-consensus published 2024
Implementation: Hanzo ACI with agent identity deployed 2025