ISO 42001 vs CSA STAR for AI: Which AI Assurance Route Should Cloud Providers Choose?

If you run a cloud platform, an AI API or a SaaS product with AI features, your enterprise customers are already asking the same question in different words: how do we know your AI is governed and secure? Two assurance routes now dominate the answer. One is ISO/IEC 42001, the international AI management system standard. The other is the Cloud Security Alliance's STAR for AI programme, built on its AI Controls Matrix.

They are often presented as rivals. They aren't. On 27 October 2026, CSA launched the full STAR for AI Level 2 programme, and its main certification routes are built on top of ISO 42001 or ISO 27001 audits. The real decision for a cloud provider is not "which one?" but "in what order, and how deep?"

This guide explains what each route proves, how they fit together, what each costs you in effort, and which path makes sense for your type of cloud or AI business.

The short answer

ISO 42001 proves you have a management system for AI: policies, roles, risk and impact assessments, objectives, internal audits and continual improvement, verified by an accredited certification body. STAR for AI proves what your AI controls actually are, at a granular, cloud-native level, and publishes that information in a public registry that buyers already use for cloud security due diligence.

ISO 42001 answers "is AI governed here?" STAR for AI answers "which specific controls protect this AI service, and who is responsible for each one?" Most serious cloud and AI providers will end up needing both. STAR for AI's Level 2 design assumes exactly that.

What ISO 42001 gives a cloud provider

ISO/IEC 42001, published in December 2023, specifies requirements for establishing, implementing, maintaining and continually improving an AI management system. It follows ISO's Harmonized Structure, so it integrates naturally with ISO 27001, which most cloud providers already hold.

For a cloud provider, the core requirements are an AI policy, defined roles, an AI risk assessment, an AI system impact assessment covering effects on individuals and society, and a Statement of Applicability across 38 Annex A controls. Those controls cover areas such as the AI system life cycle, data for AI systems, information for interested parties and third-party relationships. An accredited certification body audits the system, issues a certificate valid for three years, and returns for annual surveillance audits.

The large cloud platforms moved early. AWS was the first major cloud service provider to announce accredited ISO 42001 certification for AI services, in November 2024, covering Amazon Bedrock, Amazon Q Business, Amazon Textract and Amazon Transcribe. It completed its first surveillance audit with no findings in November 2025.

The ISO Survey counts only certificates from accredited bodies listed in IAF CertSearch, which is why announcement-based tallies run higher.

What CSA STAR for AI gives a cloud provider

CSA's STAR (Security, Trust, Assurance and Risk) programme has been the cloud industry's transparency registry for more than a decade. CSA's STAR page currently shows 4,086 registry listings from 2,876 organisations. STAR for AI extends that model to AI services.

Its foundation is the AI Controls Matrix (AICM), released in July 2025 and updated to version 1.1 in July 2026. AICM v1.1 contains 247 control objectives across 18 domains. It is built on CSA's Cloud Controls Matrix, adds AI-specific areas such as Model Security, and maps to ISO 42001, ISO 27001, the NIST AI RMF, BSI AIC4 and the EU AI Act.

What makes AICM especially useful for cloud providers is its shared-responsibility model. Every control is tagged with an owner across five roles:

• Cloud service provider: the infrastructure and platform the AI runs on
• Model provider: the organisation that trains and supplies the model
• Orchestrated service provider: the layer that connects models, tools, data and agents
• Application provider: the business that builds a product on top
• AI customer: the organisation that uses the AI service

ISO 42001 asks you to determine your role in relation to AI systems, but it doesn't split each control by supply-chain layer. AICM does, which is why it fits the way cloud AI is actually bought and built.

The control counts reflect different philosophies, not different levels of rigour. ISO 42001's 38 controls are high-level objectives you tailor through risk assessment. AICM's 247 are specific, auditable control objectives covering AI governance plus the full security stack underneath it, from identity and access management to logging, encryption and supply chain.

The STAR for AI levels and routes, explained

STAR for AI has two levels. Level 1 is about transparency. Level 2 is about independent assurance.

Valid-AI-ted is CSA's large-language-model scoring service for STAR self-assessments. It costs USD 595 for up to ten scoring attempts within a year and is free for CSA corporate members. CSA says a Valid-AI-ted AI-CAIQ submission is required for every Level 2 route, whether certification or attestation.

Adoption is still early. A September 2026 snapshot of the STAR Registry showed around 60 AI-CAIQ listings and only a handful of ISO 42001-based Level 2 entries. For a cloud provider, that means a STAR for AI listing is still a real differentiator rather than a box every competitor has already ticked.

If you are weighing these routes for your own platform, I regularly share breakdowns of ISO 42001, CSA STAR and the EU AI Act on LinkedIn. You're welcome to connect with me there and share which assurance questions your customers ask most.

ISO 42001 vs CSA STAR for AI: side by side

Neither route gives you EU AI Act conformity

Cloud providers selling into Europe often hope one of these badges will settle AI Act questions. Neither does. AICM maps to the EU AI Act and ISO 42001 lines up well with it, but only harmonised standards cited in the Official Journal give a presumption of conformity. The first, EN 18286:2026 for the quality management system of high-risk AI providers, was cited on 14 October 2026.

Which route should your cloud business choose?

A 6-step roadmap to both

1. Map your role in the AI supply chain

   For every AI service, decide whether you act as cloud provider, model provider, orchestrated service provider, application provider or more than one. This drives both your ISO 42001 scope and your AICM control ownership.

2. Publish a Level 1 AI-CAIQ

   Complete the AI-CAIQ honestly, using evidence pointers rather than marketing language. Run it through Valid-AI-ted to get scored feedback, then publish. This gives sales teams an answer immediately.

3. Build the ISO 42001 management system

   Extend your ISO 27001 ISMS with an AI policy, AI risk assessment, AI system impact assessment and a Statement of Applicability. Reuse document control, internal audit and management review.

4. Use one control library for both

   Use CSA's official AICM-to-ISO 42001 mapping so every Annex A control and AICM objective points to the same evidence. One control, one owner, one piece of evidence.

5. Choose a CSA-approved auditor and route

   Pick a certification body that can audit ISO 42001 and AICM together, or a SOC 2 auditor for the attestation route, so you avoid paying for two separate audit cycles.

6. Keep the registry listing current

   Update the AI-CAIQ when models, data sources, providers or architectures change. A stale listing undermines the transparency it was meant to show.

Common mistakes to avoid

• Treating a Level 1 self-assessment as if it were independent assurance
• Assuming ISO 42001 alone covers AI security controls such as model security and prompt-injection defences
• Claiming controls in the AI-CAIQ that belong to your model provider or customer
• Running ISO 42001 and AICM as separate projects with separate evidence
• Marketing ISO 42001 or STAR for AI as EU AI Act compliance

A prompt to choose your assurance route

Paste in your company details to get a first-draft assurance roadmap. Your compliance lead should validate it against CSA's current STAR for AI requirements.

You are an AI assurance specialist with deep knowledge of
ISO/IEC 42001:2023, ISO/IEC 27001:2022, SOC 2 and the CSA STAR
for AI programme (AI Controls Matrix v1.1 and AI-CAIQ).

Our company:
[company type: IaaS/PaaS, model provider, AI SaaS, other; AI
services offered; main markets (US, EU, other); certifications we
hold (ISO 27001, SOC 2, CSA STAR, ISO 42001); whether we provide
high-risk AI under the EU AI Act; top 5 buyer questions we get on
AI security and governance; team size for compliance]

Produce:
1. Our likely AICM role(s): cloud service provider, model
   provider, orchestrated service provider, application provider,
   AI customer, with reasons.
2. A recommended assurance route: STAR for AI Level 1, Level 2
   certification (which ISO route) or Level 2 attestation, and
   whether ISO 42001 should come first or in parallel.
3. A table of what each route would prove to buyers, and what it
   would not.
4. The existing evidence we can reuse from our current
   certifications.
5. The top 10 AICM and ISO 42001 Annex A gaps we are likely to
   have.
6. A 12-month roadmap with milestones.

Use only what I provide and mark uncertain points UNKNOWN.

ISO 42001 tells your customers that AI is governed in your organisation. STAR for AI shows them exactly how, control by control. Cloud buyers increasingly want to see both. — Kumail Mehdi

Frequently asked questions

Have a question about choosing between ISO 42001 and STAR for AI for your platform? Let's connect on LinkedIn. I'm always happy to talk AI assurance.

Sources

• CSA — STAR for AI Level 2 launch (27 October 2026)
• CSA — Launch of STAR for AI (23 October 2025)
• CSA — Understanding STAR for AI Level 2
• CSA — AI Controls Matrix v1.1
• CSA — AICM to ISO/IEC 42001 mapping
• CSA — STAR programme and registry
• CSA — STAR Registry submissions and Valid-AI-ted pricing
• CSA Labs — AICM v1.1: what changed (CCM v4.1 baseline)
• GovernanceDocs — AICM and STAR Registry snapshot (September 2026)
• ISO — ISO/IEC 42001:2023 AI management systems
• HEIC — ISO Survey 2025 results, including ISO/IEC 42001
• Atoro — How many companies are ISO 42001 certified?
• AWS — AWS achieves ISO/IEC 42001 accredited certification
• AWS — First ISO 42001 surveillance audit with no findings