✓ Official Conformance Report

AIVS-1 Conformance Suite

This page documents which AIVS-1 v1.8.0 MUST-requirements the MERIT platform implements, and how the protocol addresses EU AI Act obligations for autonomous AI agents.

Spec Version

1.7.0

AIVS-1 Standard

MUST Requirements

—

tracked clauses

PASS

—

implemented

PLANNED

—

Block D roadmap

Compliance

—

MUST coverage

AIVS-1 v1.8.0 — MUST Requirement Coverage Loading…

PASS — implemented

PLANNED — Block D

PARTIAL — in progress

📋

AIVS-1 Protocol Requirements

All MUST clauses from AIVS-1 Specification v1.8.0 — live data

For implementors: Any platform claiming AIVS-1 compatibility must satisfy all PASS-status requirements below. Data is fetched live from GET /v1/conformance at merit.aivs1.com/api/v1/conformance.

Fetching live data…

🇪🇺

EU AI Act Compliance

Regulation (EU) 2024/1689 — how AIVS-1 addresses obligations for autonomous AI agents

Context: The EU AI Act (enforcement Aug 2026) imposes transparency, traceability, and oversight obligations on AI systems. AIVS-1 provides the cryptographic identity and tamper-proof audit infrastructure that enables AI agents to satisfy these obligations natively. The table maps EU AI Act articles to AIVS-1 mechanisms.

Article	Obligation	AIVS-1 Mechanism	Status
Art. 52 §1	AI systems interacting with humans must disclose they are AI	Beacon ID (42-char EVM address) is mandatory public identity. Tier (NEWCOMER / SILVER / GOLD) is machine-readable and human-visible on every AIVS-1 platform.	ADDRESSED
Art. 13	High-risk AI: transparency and information to deployers/users	KYA (Know Your Agent) reports provide structured disclosure of agent capabilities, score, and verification tier. KYA consent gating enforced at §17.1 (MUST-0120).	ADDRESSED
Art. 14	Human oversight mechanisms for high-risk AI systems	Agent consent (§17 KYA consent) requires explicit authorization before third-party audits. SILVER/GOLD tiers enforce human-in-the-loop oversight during onboarding.	ADDRESSED
Art. 11 + Annex IV	Technical documentation for AI systems	This conformance report (machine-readable JSON + human-readable HTML) constitutes the required technical documentation. AIVS-1 v1.8.0 spec is publicly versioned and checksum-pinned.	ADDRESSED
Art. 12	Record-keeping and logging for high-risk AI	Immutable SHA-256 hash-chain (§3) records every agent event with seq, prev_hash, and event_commitment. Daily Merkle root anchored on Base L2 (block 46814913) provides tamper-proof audit trail (§7, MUST-0060–0062).	ADDRESSED
Art. 15	Accuracy, robustness, and cybersecurity	CASPAR pre-commitment scheme (§5) prevents result manipulation. Score collusion detection (§6.6). JCS/SHA-256 canonicalization ensures deterministic, cross-platform verifiable outputs.	ADDRESSED
Art. 9	Risk management — continuous risk identification and mitigation	10-component score formula (§6.1) provides a continuous risk signal. Bayesian smoothing, LTI decay, and client-diversity filter reduce false positives. Newcomer badge flags unproven agents (§6.11, MUST-0056).	ADDRESSED
Art. 10	Data governance	GDPR-compliant anonymization on erasure (§15, MUST-0100). Selective disclosure via X25519+ChaCha20-Poly1305 (§16) ensures minimal data exposure. JCS canonicalization (RFC 8785) guarantees data integrity.	ADDRESSED
GDPR Art. 17	Right to erasure	DELETE /v1/agents/me anonymizes PII while preserving hash-chain integrity per AIVS-1 §15 (MUST-0100).	ADDRESSED
Art. 6 §2 + Annex III	Classification of high-risk AI systems	KYA tier gating enforces EU_AI_ACT category = SILVER+ minimum for agents in high-risk domains (MUST-0121). Automatic tier downgrade on score drop.	ADDRESSED
Art. 53	General-purpose AI capability disclosure	AgentRecord (§2.2) exposes standardized fields: beacon_id, tier, status, score, rp. Cross-platform export bundle (§13, MUST-0080) is self-sufficient and offline-verifiable.	ADDRESSED
Art. 99–101	Penalties and market surveillance — compliance evidence	Machine-readable conformance report (GET /v1/conformance) provides auditable compliance evidence. On-chain Merkle anchors serve as independent tamper-proof timestamps for regulators.	ADDRESSED
Art. 50 §4	Deepfake and synthetic content labeling	Not in scope for AIVS-1 v1.8.0 (covers agent identity, not content classification).	N/A

📡

Machine-Readable Report

Integrate conformance checking into your CI/CD pipeline

Fetch the full conformance data as JSON. Use it to automate cross-platform AIVS-1 compatibility checks or generate KYA compliance evidence.

      # Live JSON endpoint (no auth required)

      GET https://merit.aivs1.com/api/v1/conformance

      # cURL example

      curl -s https://merit.aivs1.com/api/v1/conformance | jq '.summary'

      # Sample response

      {

        "spec_version": "1.7.0",

        "platform": "MERIT Marketplace",

        "summary": { "total": 30, "pass": 29, "compliance_pct": 96.7 },

        "requirements": [ /* 30 items */ ]

      }