AgentAudit AI
On-chain AI compliance infrastructure for autonomous AI agents. Immutable audit trails, ZK-proof verification, and automated EU AI Act reporting (Articles 9-15, 50, 53, 72) deployed on Arbitrum One and 4 other EVM mainnets. 169/169 tests passing. EU AI Act enforcement deadline August 2026.
视频
技术栈
描述
AgentAudit AI is on-chain compliance infrastructure for autonomous AI agents, built directly against the EU AI Act (Regulation 2024/1689).
Each EU AI Act article is mapped to a deployed smart contract or on-chain mechanism:
Art. 9 — Risk Management System: AuditVault + automated risk scoring (HIGH / MEDIUM / LOW) per agent action
Art. 11 — Technical Documentation (AIBOM): AIBOMRegistry.sol — CycloneDX 1.6 schema
Art. 12 — Record-Keeping: AuditVault — immutable on-chain log per action, tamper-proof by design
Art. 13 — Transparency: Every agent action timestamped and publicly readable on-chain
Art. 14 — Human Oversight: Compliance dashboard + on-chain override mechanisms
Art. 19 — Logging Requirements: Automatic per-action logging via AuditVault
Art. 26 — Deployer Obligations: AgentRegistration + KYA (Know Your Agent) fields enforced on-chain
Art. 50 — AI Content Transparency: Decision Provenance layer + on-chain logs
Art. 72 — Post-Market Monitoring: PostMarketMonitor.sol — continuous activity log
Art. 73 — Incident Reporting: IncidentRegistry.sol — automated reporting with 15/10/2 day timelines
Infrastructure:
Deployed on 5 mainnets: Mantle, Base, Arbitrum, Optimism, Polygon
ERC-8004 AgentIdentityRegistry — on-chain agent identity standard
MCP server for native agent framework integration
TypeScript SDK + ElizaOS and LangChain plugins
169/169 tests passing
Why on-chain: Off-chain logs can be deleted, altered, or lost. Blockchain-anchored audit trails cannot. For regulators, this is the difference between evidence and assertion.
Live: getagentaudit.xyz