ERC-8259: AI Agent Identity & Threat Registry

Abstract

ERC-8259 introduces a standardized interface for decentralized identity (DID) management, dynamic risk scoring, and trustless threat intelligence sharing designed specifically for autonomous AI agents operating on EVM-compatible networks.

Motivation

The transition toward an “Agentic Economy”—where autonomous AI agents negotiate, manage assets, and execute complex cross-contract logic—requires a native, programmatic trust layer. Current standards such as ERC-725 (Identity) and ERC-4337 (Account Abstraction) lack the high-frequency, verifiable risk-scoring mechanics necessary to secure A2A (Agent-to-Agent) and Agent-to-Smart-Contract interactions.

Without a global, decentralized registry to broadcast cryptographically signed threat heuristics (e.g., detecting wallet draining or mixing chain topologies), DeFi protocols are highly vulnerable to malicious autonomous actors. ERC-8259 bridges this gap by enabling protocols to programmatically query an agent’s real-time riskDelta before granting execution privileges.

Core Architecture

The standard proposes three primary interfaces:

  1. IAgentIdentity: Maps the executing wallet address to an Agent DID, storing deterministic metadata (model weight hashes, verifiable creator signatures).
  2. IAgentReputation: A stateful interface allowing designated security nodes to increment or slash an agent’s trust score based on on-chain execution behavior.
  3. IThreatRegistry: A collaborative oracle interface where standardized threat signals can be broadcasted and queried in real-time.

Discussion Points

We are submitting this draft and actively seeking community feedback on the following:

  • Sybil Resistance: Mechanisms to prevent malicious actors from spinning up transient agents to bypass the IAgentReputation slashing mechanics.
  • Threat Taxonomy: Standardizing the riskDelta integers and threat classification IDs across the EVM ecosystem.
  • Cross-Chain Propagation: Best practices for syncing the IThreatRegistry state across L2s (Arbitrum, Base, Optimism) without excessive gas overhead.

We look forward to your technical insights and rigorous feedback to help mature this standard!