NIST AI RMF 1.0 — Attestix coverage

The NIST AI Risk Management Framework 1.0 (NIST AI 100-1), released January 2023, is the most widely-referenced non-EU AI governance framework. Its four core functions — GOVERN, MAP, MEASURE, MANAGE — give organisations a vocabulary for AI risk management that US Federal agencies are required to align with under OMB Memo M-24-10. This page shows how Attestix's signed audit chains, Verifiable Credentials, and provenance records map to the AI RMF subcategories.

Why this matters

The AI RMF is voluntary guidance, not a regulation. There is no "AI RMF certified" mark. NIST itself uses the language "operationalising the AI RMF" — using it as a structuring framework for an organisation's own AI risk-management programme. Attestix is evidence tooling an organisation operationalising the AI RMF can use to gather signed evidence for specific subcategories — particularly those that touch the AI lifecycle where signed audit trails + provenance + impact-assessment records reduce manual evidence-gathering. The AI RMF Playbook + AI 600-1 Generative AI Profile are NIST's complementary operational guides; Attestix slots underneath both as the cryptographic evidence layer.

The honest coverage table

The table below covers 20 representative subcategories spanning the four functions. The complete inventory (72 subcategories across GOVERN-MAP-MEASURE-MANAGE) and the per-function tally is in §4 of the internal mapping doc.

Tally (across all 72 subcategories)

The 20 rows above are representative; for the complete count we estimate (per §4 of the internal mapping doc):

• strong-partial+: 18 (1 of which is full — MEASURE-2.8 transparency + accountability)
• partial: 18
• record-only: 22
• out-of-scope: 14 (cluster in GOVERN organisational + cultural subcategories)

Important caveat — security_check_id ships in v0.5.0. As of Attestix v0.4.0 the underlying events are emitted today, but they are NOT yet tagged with the nist.airmf.* discriminator. The v0.5.0 release registers the prefix in FRAMEWORK_REGISTRY; per-subcategory emission tagging is incremental.

What we don't do

• We do not author your organisational culture. GOVERN-4.1 (safety-first culture), GOVERN-2.1 (workforce diversity), and similar GOVERN subcategories are organisational and cultural. Any tool that claims to "verify culture" is overclaiming.
• We do not run TEVV ourselves. MEASURE-1.x and MEASURE-2.x expect your eval pipeline (Weights & Biases, MLflow, Inspect AI, garak, promptfoo). We wrap signatures around the outputs.
• We do not assess third-party risk. MAP-3.4 + CC9.2 (SOC 2) expect Vanta / Drata / UpGuard / SecurityScorecard style vendor-risk assessment. We record relationships + signed claims.
• We are not the AI RMF Playbook. NIST's AI RMF Playbook is the authoritative subcategory-by-subcategory implementation guide. We point at evidence shapes; the Playbook tells you what to do.
• We do not detect emergent behaviour. MEASURE-3.2 emergent-risk tracking points at research-grade interpretability work (Anthropic interpretability, white-box probes). We record the operator's observations.
• We do not auto-promote assertions to verifications. When a human asserts "yes, this met the criterion", the result transitions to assertion_recorded (rendered blue), NOT to passed=true (rendered green).

How to verify our coverage yourself

Python / CLI

# List every audit event tagged with an AI RMF subcategory (v0.5.0+)
attestix audit list --security-check nist.airmf.MEASURE.2.8.transparency_artefact_published

# Export a bundle scoped to AI RMF evidence (v0.5.0+)
attestix bundle export --controls nist.airmf --out my-airmf-evidence.atxbundle

# Verify chain integrity for an agent's audit log (today)
attestix verify-chain <agent-did>

JavaScript / browser

npm install attestix

import { verifyCredential } from "attestix";

const result = await verifyCredential(verifiableCheckResultJson);
// result.valid === true if the Ed25519 signature over the JCS-canonical body
// matches the issuer DID's public key.

On-chain anchor (Base L2 Sepolia testnet)

attestix anchor audit-batch --agent <did> --network base-sepolia

Mainnet schema registration is planned; testnet is the default target today.

Comparable disclosure

How other tools position themselves on AI RMF alignment.

See also

• OWASP Top 10 for Agentic Applications mapping
• ISO/IEC 42001:2023 mapping
• SOC 2 Trust Services Criteria mapping
• FRIA template (EU AI Act Art 27)
• EU AI Act compliance guide
• The internal mapping spec at attestix-cloud-plan/24-NIST-AI-RMF-MAPPING.md.

Attestix is evidence tooling for organisations operationalising the NIST AI RMF 1.0. Attestix does not issue AI RMF conformance attestations (no such attestation exists; the AI RMF is voluntary guidance), does not replace the organisation's risk-management programme, and a passing tag against a subcategory is one signal in the overall risk posture.