How to Implement ISO 42001: A Practical Guide

Reviewed by Ali Aleali, CISSP, CCSP · Last reviewed November 12, 2026

ISO 42001 is the first international standard for AI management systems. Implementing ISO 42001 means building an Artificial Intelligence Management System (AIMS) on the same Plan-Do-Check-Act structure as ISO 27001, then adding AI-specific controls in five areas:

  • AI risk and impact assessment
  • Data governance
  • Model lifecycle
  • Transparency
  • Human oversight

With an ISO 27001 or SOC 2 program in place, ISO 42001 is an extension of that program, and roughly 60-70% of the work is already done. For teams starting from scratch, the structure is the same but the path is longer.

What does ISO 42001 require?

ISO 42001 requires an AIMS that follows the same Plan-Do-Check-Act model as ISO 27001. The core requirements:

  • Leadership commitment and AI policy. Management defines the organization's approach to responsible AI, including ethical principles, risk appetite, and accountability structures.
  • AI risk assessment process. Separate from the information security risk assessment. AI risks include biased outputs, poor data quality, lack of explainability, model drift, and unintended results of autonomous decisions.
  • AI Impact Assessment. Evaluates the effects of AI systems on individuals, groups, and society, beyond technical risk.
  • Annex A controls. 38 AI-specific controls, covering areas such as AI policy, the AI system lifecycle, data management, and third-party relationships.
  • Annex B guidance. Implementation guidance for each Annex A control.
  • Continual improvement. Monitoring, measurement, internal audit, and management review, same as any ISO management system.

Key Insight

The AIMS mirrors the ISO 27001 ISMS. Teams running an ISMS apply a familiar structure to a new domain, AI governance, without learning a new framework.

For a deeper dive into what ISO 42001 is and why it matters, see our overview.

Not Sure Where Your AI Governance Stands?

Take the ISO 42001 Readiness Scorecard. Fifteen questions, five minutes. You get a snapshot of your current AI governance maturity, and we get enough context to make a strategy call useful.

Take the Scorecard

How is ISO 42001 different from ISO 27001 and SOC 2?

ISO 27001 and SOC 2 teams already know how to run a management system. ISO 42001 adds five areas that security programs do not cover.

1. AI Risk and Impact Assessments

Security risk assessments focus on the confidentiality, integrity, and availability of data. AI risk assessments ask what harm the AI system's outputs can cause:

  • Bias and fairness: Does the model produce discriminatory outcomes for certain groups?
  • Explainability: Can you explain how the model reached a specific decision? Can the people affected by that decision understand the explanation?
  • Model drift: Is the model's accuracy degrading over time as real-world data shifts away from training data?
  • Autonomous decision scope: Where does the AI system make decisions without human review, and what's the risk if those decisions are wrong?

Run AI risk assessments during development, before deployment, and periodically in operation.

2. Data Governance Beyond Security

ISO 27001 covers data protection: encryption, access control, and retention. ISO 42001 adds AI data governance:

  • Data provenance: Where did the training data come from? Is the sourcing documented and auditable?
  • Data quality: Is the training data representative, accurate, and complete? What processes exist to verify this?
  • Bias in data: Has the training data been evaluated for inherent biases that could propagate through model outputs?
  • Data lifecycle: How is training data managed, updated, and retired as models evolve?

With third-party models (OpenAI, Anthropic, open source), you may not control the training data. You still document what you know about the data and assess the risk of what you do not know.

Third-Party Model Governance

The standard doesn't require you to train your own models. It requires you to govern how you use them. For companies integrating third-party AI via APIs, that means documenting model selection criteria, monitoring outputs, and maintaining fallback procedures when upstream models change.

3. Model Lifecycle Management

AI models need their own lifecycle. ISO 42001 requires documented processes for each stage:

MODEL LIFECYCLE STAGES

Development

How models are designed, trained, validated, and tested before deployment.

Deployment

How models move from development to production, including approval gates.

Monitoring

How model performance, accuracy, and behavior are tracked in production.

Updating

How models are retrained, fine-tuned, or replaced when performance degrades.

Retirement

How models are decommissioned, including what happens to their outputs and dependent systems.

Teams using third-party AI (APIs from OpenAI, Anthropic, or Google, or open-source models) also define how they monitor and manage models they did not build.

4. Transparency and Explainability

AI systems must give stakeholders appropriate transparency. Appropriate depends on the system's risk level, but at minimum:

  • Users should know they're interacting with an AI system
  • Affected parties should have access to information about how AI decisions are made
  • The organization should be able to explain AI system behavior to regulators, auditors, and customers

A privacy policy is not enough. Transparency needs technical documentation of how models work, what inputs the models use, and how outputs are generated, written for each audience.

5. Human Oversight

ISO 42001 requires organizations to define:

  • Which AI decisions require human review before action
  • What escalation paths exist when the AI system produces unexpected or potentially harmful outputs
  • How human oversight is documented as evidence

How do you implement ISO 42001 on an existing ISO 27001 or SOC 2 program?

Extend the existing program. Do not start over. The management system, policies, risk framework, internal audit process, and GRC platform configuration all transfer.

Head Start from ISO 27001 or SOC 2

Companies with an existing management system can expect 60-70% of ISO 42001 requirements to be already addressed. The remaining effort focuses on AI-specific controls, risk assessments, and governance processes.

What you add:

  • AI-specific policies (AI ethics, acceptable use of AI, AI risk management)
  • AI risk and impact assessment process and templates
  • Annex A controls mapped to your AI systems
  • Data governance procedures specific to AI (provenance, quality, bias evaluation)
  • Model lifecycle documentation
  • Transparency and human oversight controls
  • AI-specific sections in your Security Program Manual

Implementation follows three phases:

IMPLEMENTATION PHASES

Assess

Inventory your AI systems. Identify which ones fall under ISO 42001 scope. Conduct AI risk and impact assessments. Map existing controls (from ISO 27001 or SOC 2) to ISO 42001 requirements and identify the gaps. The gap is usually in the five AI-specific areas (risk and impact assessments, AI data governance, model lifecycle, transparency, and human oversight), not in the management system fundamentals.

Build

Close the AI-specific gaps. Write AI policies. Design the AI risk assessment process. Document model lifecycle management. Configure the GRC platform with ISO 42001 controls (Drata and Vanta both support ISO 42001). Build the evidence architecture for AI-specific controls.

Operate

Run the AIMS on a cadence: AI risk assessments at planned intervals and whenever an AI system changes, continuous model monitoring, periodic data quality reviews, and oversight records for every reviewable decision. The coordinator role and feedback loops from ISO 27001 or SOC 2 carry over.

Companies without an in-house security lead can hand that cadence to a vCISO. Truvo's vCISO services operate the AIMS on the same program as ISO 27001 or SOC 2.

Timeline: 6-12 months from scratch depending on scope and complexity, similar to a first ISO 27001 certification, or 2-4 months for the AI-specific additions on an existing program.

What should ISO 42001 implementation services deliver?

A complete engagement builds the AIMS and hands over an audit-ready evidence set before the certification body begins Stage 1. Deliverables:

  • AI system inventory, with a scope statement and a classification for each system
  • AI risk assessment method and AI risk register
  • AI system impact assessments for every in-scope system
  • AI policy plus supporting policies (acceptable use of AI, data governance for AI, model lifecycle)
  • Statement of Applicability covering the 38 Annex A controls, mapped to existing ISO 27001 or SOC 2 controls
  • Evidence collection configured in the GRC platform, where one is in use
  • A completed internal audit and management review before the certification audit
  • Stage 1 and Stage 2 audit support with the certification body

Three checks separate providers:

  • Independence from the certification body. ISO/IEC 17021-1 bars a certification body from consulting on the management system it certifies, so the firm that builds the AIMS and the firm that audits the AIMS are always two separate firms.
  • Experience with the AI systems in scope. The provider should be able to assess third-party model APIs, fine-tuned models and agent workflows as well as write policy.
  • Reuse of the existing program. With ISO 27001 or SOC 2 in place, the provider maps existing controls first and quotes only the AI-specific gap.

Consulting, audit and platform price ranges are in the ISO 42001 certification cost breakdown. Truvo's ISO 42001 implementation services cover all three phases, from the AI system inventory through to the certification audit.

What are the ISO 42001 gaps in 2026?

The standard is ahead of the market. ISO 42001 was published in December 2023. As of 2026, relatively few companies are certified and certification bodies are still building expertise, so early adopters face some ambiguity in how controls are interpreted and evidenced.

Standards are lagging technology by about 3 years. Large language models, agent architectures, the Model Context Protocol, and open-source model hosting raise questions ISO 42001 does not address directly yet. Teams apply the standard's principles to technology the authors did not fully anticipate.

GRC platform support is emerging. Major platforms now include ISO 42001 control frameworks, but automated testing and evidence collection for AI-specific controls are still early. Expect more manual evidence than with ISO 27001 or SOC 2 on the same platform. For a detailed comparison of platform capabilities, see our ISO 42001 compliance software review and cost benchmarking guide.

Strategic Value of Early Certification

Early certification is an advantage, particularly in regulated industries and enterprise sales. As EU AI Act deadlines approach, ISO 42001 gives a defensible governance posture before the market requires one.

ISO 42001 Certification-Ready in 12 Weeks

Our ISO 42001 Accelerator covers AI risk assessment, AIMS implementation, GRC platform setup, and certification body coordination. Fixed price, no scope creep.

Not ready to talk yet? Take the readiness scorecard first.

Frequently Asked Questions

How do you implement ISO 42001?

In three phases. Assess: inventory AI systems, run AI risk and impact assessments, and find gaps against Annex A. Build: write AI policies, design data governance and model lifecycle processes, and configure the GRC platform. Operate: run the AIMS on a cadence with continuous monitoring and periodic reviews. Companies with ISO 27001 extend the existing management system.

What are the ISO 42001 requirements?

An AIMS with leadership commitment and an AI policy, AI risk and impact assessments, the 38 Annex A controls, and continual improvement through monitoring, internal audit, and management review, on the same Plan-Do-Check-Act model as ISO 27001.

How long does ISO 42001 certification take?

6-12 months from scratch, similar to a first ISO 27001 certification. The timeline depends on the number of AI systems in scope, program maturity, and team availability.

What is the difference between ISO 42001 and ISO 27001?

ISO 27001 covers information security management. ISO 42001 covers AI management. Both use Plan-Do-Check-Act, and ISO 42001 adds AI risk and impact assessments, AI data governance, model lifecycle management, transparency, and human oversight. See ISO 42001 vs ISO 27001.

Do I need ISO 27001 before getting ISO 42001?

No. ISO 42001 is a standalone standard, but companies with ISO 27001 have a large head start because the management system transfers.

How long does ISO 42001 implementation take if we already have ISO 27001 or SOC 2?

About 2-4 months for the AI-specific additions, because roughly 60-70% of the management-system work is already done. The added work covers AI risk and impact assessment, data governance, model lifecycle, transparency, and human oversight.

Does ISO 42001 certification make us compliant with the EU AI Act?

No. ISO 42001 maps to the EU AI Act's core requirements for risk management, transparency, data governance, and human oversight, but the mapping is not one-to-one, and certification alone does not meet the Act's obligations for high-risk AI systems. See the ISO 42001 and the EU AI Act breakdown.

Should we use Drata or Vanta for ISO 42001?

Both offer ISO 42001 support with automated evidence collection, cross-mapping to ISO 27001, and AI risk workflows. The right choice depends on your stack, AI use cases, and team. See the Drata vs Vanta for ISO 42001 comparison.

How much does ISO 42001 compliance software cost?

Pricing varies with the GRC tool, implementation fees, and audit costs. The ISO 42001 software cost benchmarking breaks total cost into platform, implementation, and certification.

Contact Us

Share this article:

ISO 27001 vs SOC 2 Control Mapping Spreadsheet

All 93 Annex A controls mapped to SOC 2, and back. Enter your email to open the Google Sheet.

About the Author

Former security architect for Bank of Canada and Payments Canada. 20+ years building compliance programs for critical infrastructure.