ISO 42001 preparation for your 2027 AI plans.

Avantcert’s ISO 42001 reference package prices are $4,000 for a small team, $8,000 for a medium team and $12,000 for a large team. The package includes preparation and the initial independent ISO 42001 certification audit for the agreed scope. Your final fee depends on team size and scope. We negotiate it on a call.

  • Already hold ISO 27001? Your management system carries over, so most of what is left is the AI-specific work.
  • All 38 Annex A controls, your AI policies and the Statement of Applicability, written for your systems.
  • Your certificate comes from an independent body accredited for ISO/IEC 42001, such as a UKAS- or ANAB-accredited body.
Organizations supported by Avantcert
3,000+
Annex A controls written for you
38
Engineering hours for an API deployer
~6
Step 1 of 2

Instant ISO 42001 scope and cost estimate

How do you use AI?
What do you hold today?

You see the estimate before we ask for an email.

Don't over-scope your audit.

Scope sized to your AI role

We draw your AIMS boundary around how you actually use AI. A company calling an LLM API does not need the model-training evidence a lab does, and the auditor will not expect it.

LLM API and RAG deployers

You call models from OpenAI, Anthropic or Google, or run RAG, copilots and agents on top of them.

Where the work is
A.10 supplier controls, A.9 responsible use, and how prompt and output data are handled.
Typical timeline
6–8 weeks with ISO 27001 in place

Model fine-tuners

You fine-tune or adapt third-party models on your own data or your customers' data.

Where the work is
A.7 data provenance and quality, plus bias and performance testing under A.6.2.4.
Typical timeline
8–10 weeks with ISO 27001 in place

Model producers

You train foundation or proprietary models and ship them inside your product.

Where the work is
The full A.6 life cycle from requirements to monitoring, plus A.7 data controls.
Typical timeline
10–12 weeks with ISO 27001 in place
If you hold ISO 27001

Why ISO 27001 holders finish sooner

ISO/IEC 42001 uses the same harmonized structure as ISO/IEC 27001, and its Annex D covers running the two as one integrated management system.

Clause 4ContextExtendAdd your AI systems and your role (deployer, developer) to the scope.
Clause 5LeadershipExtendAdd an AI policy and AI roles to the leadership you already have.
Clause 6PlanningNet-new AI work6.1.2 AI risk assessment, 6.1.3 risk treatment and 6.1.4 AI system impact assessment.
Clause 7SupportReuse as isDocument control and awareness carry over; competence extends to AI roles.
Clause 8OperationNet-new AI workRun the AI risk and impact assessments (8.2–8.4) for each system.
Clause 9EvaluationReuse as isInternal audit and management review carry over with AI inputs added.
Clause 10ImprovementReuse as isNonconformity and corrective action processes are shared.
Annex AAnnex A.2–A.10Net-new AI work38 AI controls: policy, impact, life cycle, data, users, suppliers.
  • Reuse as is
  • Extend for AI
  • Net-new AI work, written by us

SOC 2 holders reuse controls and evidence but not the management system clauses, so their timeline sits between the two.

What you get

Look inside the 6 documents we write for you

Each one is written for your systems and reviewed with you before the auditor sees it. The excerpts below are illustrative; client documents stay confidential.

AI system inventory

Every AI system in scope, including third-party models inside features nobody catalogued, with owner, purpose and risk tier.

Meets Clause 4.3 scope, A.4 resources

AIMS-INV-01v1.2 · Owner AI Governance Lead
AI system inventory
Organization: Client name withheld
SystemPurposeAI roleModelRisk tier
Support reply copilotDrafts replies to ticketsDeployerThird-party LLM APIMedium
Contract clause searchRAG over customer contractsDeployerEmbeddings + LLM APIMedium
Lead scoring modelRanks inbound leadsDeveloperIn-house gradient boostingLow
Illustrative excerpt. Fictional systems and figures.
Software vs a team

Why Vanta or Drata alone won't get you through ISO 42001

Compliance platforms track controls and collect evidence well, but your team still has to write the impact assessments, the AI inventory and the reasoning behind each control decision.

Compliance software alone compared with Avantcert for ISO 42001
ISO 42001 workCompliance software aloneAvantcert
AI system inventoryA template your team fills inBuilt with your team in one working session
Clause 6.1.4 impact assessmentsA generic template per systemWritten per system and reviewed with you
Statement of ApplicabilityA list of 38 controls to complete38 decisions, each with its justification
AI policiesTemplates you editWritten for how your products use AI
EvidenceAutomated for connected cloud systemsGathered by us, inside your platform if you have one
Internal auditYou arrange itWe run it before the certification audit
Stage 1 and Stage 2Your team faces the auditorWe sit in both audits with you
Engineering timeWeeks of reading the standard and writingAbout 6 hours for an API deployer

Vanta and Drata are compliance automation platforms; this compares their self-serve model with a done-for-you service. Many of our clients keep their platform, and we work inside it.

Week by week

The roadmap, with the hours we need from your engineers

A typical plan for an LLM API deployer that already holds ISO 27001. Certification-body scheduling can add a week or two.

  1. Week 1

    Scoping and AI inventory

    One call to agree the AIMS boundary and walk through the AI systems you run, including the ones inside vendor features.

    1.5 hrs
  2. Weeks 2–4

    We write the AIMS

    AI policy, risk and impact assessments, the Statement of Applicability and the vendor packs. You review two sets of drafts.

    2 hrs
  3. Week 5

    Internal audit and management review

    We audit your AIMS the way the certification body will, then close anything we find before booking the audit.

    1 hr
  4. Weeks 6–8 target

    Readiness and audit coordination

    Confirm readiness and agree Stage 1 and Stage 2 dates with an independent accredited body. Audit availability and corrective actions can extend the schedule.

    1.5 hrs

Total engineering time: about 6 hours. The certificate is valid for three years, with a surveillance audit in each of years two and three.

Regulation and procurement

How ISO 42001 lines up with the EU AI Act and enterprise buyers

ISO 42001 is voluntary, and certifying to it does not make you compliant with the EU AI Act. It does build most of the governance evidence the Act asks of high-risk AI providers, and it is the certificate buyers recognize.

EU AI Act

Applies to anyone placing AI systems on the EU market, wherever the company is based. Annex III high-risk obligations start on 2 December 2027 under Regulation (EU) 2026/1744; prohibited practices, general-purpose AI duties and Article 50 transparency already apply.

Enterprise security reviews

Security questionnaires now carry AI sections: which models you use, how customer data reaches them, who approves changes. A certificate, or a dated plan to one, answers them once instead of deal by deal.

A certificate buyers recognize

ISO/IEC 42001 is an international standard. Certification bodies accredited by national bodies such as UKAS and ANAB audit it under the ISO/IEC 42006 rules, and a buyer in any country can check your certificate with the issuing body.

EU AI Act high-risk requirements, mapped to ISO/IEC 42001

EU AI Act high-risk requirements mapped to ISO/IEC 42001 clauses and controls
EU AI Act requirementISO/IEC 42001What we write
Art. 9 Risk management systemClauses 6.1.2–6.1.4, 8.2–8.4AI risk and impact assessments
Art. 10 Data and data governanceA.7 Data for AI systemsData provenance and quality records
Art. 11 and Annex IV Technical documentationA.6.2.3, A.6.2.7Design and technical documentation
Art. 12 Record-keepingA.6.2.8 Event logsLogging requirements per system
Art. 13 Transparency to deployersA.8 Information for interested partiesUser and customer documentation
Art. 14 Human oversightA.9 Use of AI systems, impact assessmentOversight measures per system
Art. 15 Accuracy and robustnessA.6.2.4 Verification and validationTest and monitoring evidence
Art. 17 Quality management systemClauses 4–10The AIMS itself

Presumption of conformity with the AI Act comes from harmonized European standards once they are published. ISO 42001 certification does not provide it, but the same documents serve both.

Questions buyers ask before they start

How long does ISO 42001 certification take?

Plan six to sixteen weeks for preparation, depending on scope and evidence readiness. An LLM API deployer that already holds ISO 27001 typically finishes in six to eight weeks; a company training its own models takes ten to twelve. Starting without ISO 27001 adds about four weeks. These are preparation planning estimates. Certification-body availability and corrective actions determine the final certification date.

How much does ISO 42001 certification cost?

Avantcert’s reference package prices are $4,000 for a small team, $8,000 for a medium team and $12,000 for a large team. The package includes preparation and the initial independent ISO 42001 certification audit for the agreed scope. Your final fee depends on team size and scope. We negotiate it on a call.

We only call OpenAI or Anthropic APIs. Do we need the whole standard?

You need every clause, but Annex A is applied in proportion to your role. A deployer's weight sits in supplier controls, responsible use and data handling, and controls that don't apply are excluded with a written justification in the Statement of Applicability.

Will this pull our engineers off the roadmap?

For an API deployer with ISO 27001 we plan on about six hours of engineering time across the project: one scoping call, two document reviews, a management review and the audit interviews. We write everything else.

We already hold ISO 27001. What is actually left to do?

Clauses 4 to 10 carry over with AI added to scope, policy and review. The net-new work is the AI risk assessment, the AI system impact assessment and the 38 Annex A controls. That is the part we write.

Can you guarantee we pass the audit?

No consultancy can promise a certification body's decision. We don't book Stage 1 until our internal audit shows your AIMS is ready, and if the auditor raises a finding, we work through the corrective action with you.

Who issues the ISO 42001 certificate?

An independent certification body accredited for ISO/IEC 42001, which audits under the ISO/IEC 42006 rules. Avantcert prepares you and never audits you, because the organization that builds a management system cannot also certify it.

Is ISO 42001 required by law?

No. It is a voluntary standard. Buyers ask for it in security reviews and tenders, and it builds much of the governance evidence the EU AI Act asks of high-risk AI providers, but certification does not by itself make you compliant with the Act.

Can we keep using Vanta or Drata?

Yes. Many clients keep their platform for evidence collection, and we write the ISO 42001 documents and work inside it.

How long is an ISO 42001 certificate valid?

Three years. There are surveillance audits in years two and three, then recertification. Impact assessments are revisited when an AI system changes materially.

Agree your ISO 42001 scope and final fee on a call

Two questions now, then a dated plan from a consultant within 24 hours.

How do you use AI?
What do you hold today?
Sample · Clause 6.1.4

AI system impact assessment

Sample for illustration. The system and organization are fictional.

AI system
Support reply copilot: drafts replies to customer support tickets
AI role
Deployer, using a third-party LLM through an API
Intended purpose
Reduce first-response time; a human agent sends every reply
Data processed
Ticket text, which can include customer names, account details and complaints
Foreseeable misuse
An agent sends a draft without reading it; the model is asked for legal or financial advice
Assessed impacts
AffectedPotential impactSeverity / likelihoodMitigation
Individual customersActs on an incorrect answer and loses moneyMedium / LowAgent approval on every reply; refusal rules for legal and financial topics
Non-English speakersReceive lower-quality replies than othersMedium / MediumQuality tested per language each month; threshold of a 5-point gap
Support agentsOver-rely on drafts and lose product knowledgeLow / MediumQuarterly review of 50 sampled replies
Human oversight
Agent reviews and sends every reply; agents can flag a draft as harmful
Residual impact
Low, accepted by the AI governance lead
Review trigger
Model or provider change, new data category, or 12 months