Skip to main content
Attestix
Security

ISO/IEC 42001:2023 (AI Management System) — Attestix coverage

How Attestix's signed audit chains, Verifiable Credentials, and provenance records map to the ISO/IEC 42001:2023 Annex A controls and AIMS process clauses. Honest per-control coverage — Attestix is evidence tooling, not an AI Management System.

ISO/IEC 42001:2023Information technology — Artificial intelligence — Management system — is the first international management-system standard specifically for AI. Published in December 2023, it defines how organisations establish, implement, maintain, and continually improve an AI management system (AIMS). This page shows how Attestix's signed audit chains, Verifiable Credentials, and provenance records map to ISO/IEC 42001's Annex A controls and process clauses.

Why this matters

ISO/IEC 42001 is a management-system standard, not a product-conformance standard. Like ISO/IEC 27001 (information security), conformance is awarded to an organisation by an accredited certification body after a defined audit programme. Attestix does not issue ISO 42001 certifications and is not a certification body. What Attestix ships is the cryptographic evidence layer an AIMS owner can use during audit preparation — particularly for the Annex A controls that touch the AI lifecycle (data, model, deployment, monitoring) where signed audit trails reduce manual evidence-gathering by weeks.

The honest coverage table

The table below covers 14 representative rows spanning 28 of the 38 Annex A controls plus the AIMS process clauses (Clauses 4–10). The complete control inventory and ratings are in the internal mapping doc.

ControlMitigation surface in AttestixCoverageEvidence shape
Clauses 4-5 Context + Leadership — AIMS scope, AI policy, leadership commitmentNone directly. Attestix records signed attestations that policy + scope exist; AIMS culture is organisational.record-onlyProviderAssertionCredential
Clauses 6-8 Planning + Support + Operation — risk treatment, documented information, operational changeProfile-based intended-purpose declaration; hash-chained documented-information changes; cross-walks to v0.5 EU AI Act Art 9 risk-management checkspartialAudit chain + VerifiableCheckResult
A.2 Policies for AI — AI policy, alignment with other policies, reviewOperator-uploaded policy fingerprint + signed attestation; we do not author policyrecord-onlyProviderAssertionCredential + bundle-asset hash
A.3 Internal organization — AI roles + responsibilities, concern reportingDID-based per-role identity; concern reports auditable via record_action(entry_type="concern_report")partialProviderAssertionCredential + audit chain
A.4 Resources for AI systems — data, tooling, compute, humanTraining-data fingerprint, model-lineage records, compute envelope declarations, per-tool invocation logstrong-partialHash-chained provenance + VerifiableCheckResult
A.5.2-5.5 Impact assessment — process + documentation + individual + societalFRIA template (see FRIA page); deterministic completeness checks; never auto-promote assertion_recorded to passed=truestrong-partialImpactAssessmentCredential + VerifiableCheckResult
A.5.6 Environmental impact — energy + resource considerationsOperator-declared training energy (kWh) / inference energy (Wh) / water-cooling impact; we record, we do not measurerecord-onlyProviderAssertionCredential
A.6.1.2-1.3 Responsible AI objectives + design processSigned intended_purpose + responsible_ai_objectives field; we record, we do not judgerecord-onlyProviderAssertionCredential + audit chain
A.6.1.4 Requirements + specificationsStructured system_specifications field; dual-tagged with EU AI Act Art 11 / Annex IV §1strong-partialProviderAssertionCredential + dual-tagged VerifiableCheckResult
A.6.2 Lifecycle — requirements + design + V&V + deployment + operation + technical docs + event logsPer-stage provenance entries; audit chain IS the A.6.2.8 event log; cross-walk to EU AI Act Arts 12 + 15strong-partialAudit chain + per-stage VerifiableCheckResult
A.7 Data — training + acquisition + quality + provenance + preparationTraining-data fingerprint with source + license + acquisition-date; operator-declared quality metrics; structured preparation log; cross-walk to EU AI Act Art 10strong-partialHash-chained provenance + ProviderAssertionCredential
A.8 Information for interested parties — system documentation + external reports + incidents + transparencyA2A-shaped signed agent cards; v0.5 incident-reporting collection (Art 73 cross-walk); existing generate_declaration_of_conformitystrong-partialAgentIdentityCredential + IncidentReportCredential + EUAIActComplianceCredential
A.9 Use of AI systems — responsible-use process + objectives + intended useintended_purpose profile + UCAN delegation chains binding deployer-side capabilitystrong-partialProviderAssertionCredential + UCAN tokens
A.10 Third-party + customer relationships — responsibility allocation + suppliers + customersUCAN capability allocation across supplier/customer DIDs; third-party AgentIdentityCredential acceptancepartialUCAN chain + supplier AgentIdentityCredential

Tally

  • full: 0 (no control is fully satisfied by a tool — ISO 42001 is a management-system standard; no tool is an AIMS)
  • strong-partial: 6 (data, lifecycle, impact assessment, transparency, deployer use, resources — the technical lifecycle clusters)
  • partial: 3 (planning/support/operation, internal organization, third-party)
  • record-only: 5 (context/leadership, policies, environment, objectives + processes)
  • out-of-scope: 0 (every control family has at least record-only Attestix surface)

Important caveat — security_check_id ships in v0.5.0. As of Attestix v0.4.0 the underlying events listed above are emitted today, but they are NOT yet tagged with the iso42001.* discriminator. The v0.5.0 release registers the prefix in the FRAMEWORK_REGISTRY closed set; per-control emission tagging is incremental. Until then you can already query the underlying events by action and reconstruct the same coverage manually.

What we don't do

  • We do not author your AI policy. A.2 policy controls expect the AIMS owner's documented policy. We record a signed fingerprint that the policy exists; we never write it.
  • We do not measure environmental impact. A.5.6 expects the operator's energy + resource disclosure. Tools like CodeCarbon and ML CO2 measure; we sign the measurements.
  • We do not assess substantive content of impact assessments. Per the v0.5 anti-compliance-theater design, when a deployer asserts a FRIA section is complete + appropriate, the result transitions to assertion_recorded (rendered blue), NOT to passed=true (rendered green). Attestix never claims to have verified the substance of a human assertion.
  • We are not your Statement of Applicability (SoA) authoring tool. The SoA is an AIMS-owner artefact. We offer a bundle export --controls iso42001 flag (v0.5) so the AIMS owner can include Attestix evidence in their SoA; we do not author the SoA.
  • We do not run third-party audits. ISO 42001 certification is awarded by accredited certification bodies (ANAB-, UKAS-, JAS-ANZ-accredited). We provide audit-ready evidence; the audit + certification are separate organisational journeys.

How to verify our coverage yourself

Python / CLI

# List every audit event tagged with an ISO 42001 control (v0.5.0+)
attestix audit list --security-check iso42001.A_7_4.data_quality_attestation_recorded

# Export a bundle scoped to ISO 42001 evidence for your SoA (v0.5.0+)
attestix bundle export --controls iso42001 --out my-aims-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(impactAssessmentCredentialJson);
// 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 ISO 42001 alignment. We name approaches found through public documentation; we do not disparage.

ToolStated ISO 42001 positionWhere to read more
Microsoft Agent Governance ToolkitPublishes docs/compliance/iso-42001-mapping.md mapping toolkit features to Annex A controls; CloudEvents-to-Azure-Monitor evidence path; SPIFFE identity (different shape from W3C DID)github.com/microsoft/agent-governance-toolkit
VantaEnd-to-end ISO 42001 readiness platform — policy authoring + employee training + evidence collection + auditor coordination + continuous monitoring dashboards. Complementary to Attestix; we slot underneath as the cryptographic evidence layervanta.com
DrataSame posture as Vanta — GRC platform with ISO 42001 module; we slot underneathdrata.com
SecureframeSame posture as Vanta + Dratasecureframe.com
ISO 42001 certification bodies (BSI, TÜV Rheinland, DNV, LRQA, etc.)The actual ISO 42001 certification awarders. Attestix is evidence input to their audit, not a competitoriaf.nu/iaf-mra/ for accredited bodies

See also


Attestix is evidence tooling for organisations operating an ISO/IEC 42001 AIMS. Attestix is not an AI Management System, is not an ISO 42001 certification body, and a passing tag against an Annex A control ID is one signal in the organisation's overall conformity posture — not a conformance verdict. The organisation's AIMS owner + the accredited certification body remain the authoritative voices.