Data Classification Tools: Top Vendors Compared in 2026
Key Takeaways: Data classification is the foundation on which modern data...
ISO/IEC 42001 at a glance
ISO/IEC 42001 is the international standard for an artificial intelligence management system, or AIMS. ISO and IEC published it in December 2023, making it the first standard of its kind. It applies to organizations that develop, provide, or use AI systems. Most importantly, ISO 4200 is certifiable, which means that an accredited third party can audit you against it and issue a certificate.
What does ISO 42001 require? A documented management system for AI: defined accountability, risk and impact assessment, data governance, life cycle controls, transparency to affected parties, supplier oversight, and continuous monitoring.
Is ISO 42001 certifiable? Yes. It’s a Type-A management system standard. Certification bodies audit your management system and issue a certificate with a defined scope.
Does ISO 42001 certify an AI model? No. It certifies the system your organization uses to govern AI, rather than approving any specific model, product, or vendor.
Does ISO 42001 satisfy the EU AI Act? No. The standard is voluntary and has no presumption of conformity. The two overlap heavily, and the documentation you build for 42001 feeds directly into what the Act requires.
Who should pursue it? Companies selling AI-enabled products into regulated markets, enterprises deploying AI internally at scale, and vendors whose deals keep stalling in security review over AI governance questions.
Inside the standard
ISO 42001 follows the same harmonized structure as ISO 27001 and ISO 9001. Clauses 4 through 10 cover organizational context, leadership, planning, support, operation, performance evaluation, and improvement, wrapped in a Plan-Do-Check-Act (PDCA) Cycle.
If your team has been through a 27001 audit, much of it will look familiar. ISO maintains its own plain-language explainer of the standard for teams starting from scratch.
Most controls you need to know about live in Annex A, which lists 38 controls across nine objectives:
- AI policy
- Internal organization and assigned responsibility
- Resources for AI systems
- AI system impact assessment
- AI system life cycle management
- Data for AI systems
- Information available to interested parties
- Responsible use of AI systems
- Third-party and customer relationships
Annex B provides implementation guidance for each control. Annex C catalogs AI-specific objectives and risk sources. Annex D addresses application across sectors and domains.
How is ISO 42001 different from ISO 27001?
With the structural overlap between the two standards, it’s tempting to view 42001 as 27001 with a few AI paragraphs added on. But that would be missing the point.
Information security risk assessment asks what could happen to the organization. AI impact assessment asks what could happen to the people on the other end of the system: candidates screened by a model, customers scored by one, patients triaged by one. Teams doing this for the first time tend to underestimate how much work that shift creates.
The standard also spans the AI value chain, so a company training its own models and a company that bought a Microsoft Copilot license both fall in scope, with the controls weighted differently for each.
Data runs through it all. Provenance, quality, lineage, and appropriate use sit at the center of the control set. An auditor will ask what data feeds your AI systems, who approved that use, and how you know sensitive content stayed out of it.
Very few organizations can answer without a substantial data discovery process.
The rest of the ISO/IEC 42000 series
ISO/IEC 42006:2025, published in July 2025, sets requirements for the certification bodies that audit against 42001. It supplements ISO/IEC 17021-1 with AI-specific competence and audit-time rules. Accreditation bodies including ANAB, and the Standards Council of Canada now run AIMS accreditation schemes against it.
If you’re evaluating a vendor’s certificate, ask whether the issuing body holds 42006 accreditation. Plenty of certificates in circulation came from bodies that lack it.
ISO/IEC 42005:2025 covers AI system impact assessment methodology, giving structure to one of the harder Annex A requirements.
ISO/IEC 23894:2023 provides AI risk management guidance mapped to ISO 31000.
ISO/IEC 22989 supplies the terminology baseline that the rest of the series builds on.
How is ISO 42001 different from the EU AI Act?
ISO 42001 is a voluntary international standard. The EU AI Act is enforceable law, with additional high-risk system obligations that will soon take effect.
The harmonized standards that confer presumption of conformity are being developed through CEN-CENELEC JTC 21, separately from ISO. Certification against 42001 will do nothing for you legally.
The practical overlap is another matter. Risk assessments, impact assessments, life cycle records, and supplier controls built for an AIMS map onto the technical documentation the Act expects.
Organizations already certified to 42001 start their AI Act readiness work with most of the governance foundations already standing.
What certification asks you to prove
Audits rarely fail because a policy document is missing; rather, they fail when the organization can’t show that the policy was enforced.
For AI, that gets uncomfortable fast.
Say your AIMS scope lays out that employees use approved AI tools under defined conditions, and the auditor asks you to demonstrate it. For example, which tools are running across the organization right now ? Or, what data has moved into them, and who signed off on those flows? And if someone broke the policy, what happened next, and where is that written down?
Most companies can answer the first question easily. After that, though, the answers get vague. The tools employees signed up for on their own rarely show up, and neither does the sensitive content that traveled into them.
We put “treating governance as a paper exercise” at the top of our list of the most common AI governance mistakes enterprises make. An ISO 42001 audit asks for the records a working policy leaves behind.
The data problem that underpins every AIMS
Read the nine control objectives closely and you’ll see that they all point to the same two questions. What data is involved, and who decided that use was acceptable? Impact assessment depends on the answer. And so does responsible use, which in practice comes down to what employees paste into a chat window when they are behind on a deadline. Supplier controls ask the same two questions again, this time directed at what leaves your environment and who holds it afterward.
Every one of these controls is a data governance issue before it is an AI issue.
Organizations that treat ISO 42001 as a documentation exercise end up with a management system describing an imaginary company, while the real one runs on data nobody’s mapped.
How Concentric AI supports ISO 42001 visibility and auditing
Concentric AI works on the evidence layer of an AIMS: what data exists, where it travels, which AI tools consume it, and what the organization did about it. That begins with autonomous discovery and categorization of business-critical and compliance-sensitive data across cloud, on-premises, and SaaS environments.
Shadow AI discovery
Concentric AI identifies AI tools in use across the organization, including the ones nobody registered with IT. Free-tier accounts, browser extensions, personal logins on corporate devices, and departmental tools adopted without review all become visible.
Before an AIMS scope statement can hold up, you have to know what falls inside it.
Read our guide to the hidden risks of shadow AI to see how far this typically extends inside an average enterprise.
Prompt and response monitoring shows what employees are sending to these tools, with inline categorization applied to the content itself rather than to file names or storage locations.
Enforcement follows with warn, block, or redact actions, so the control produces an outcome you can point to later.
Enterprise AI platforms
For sanctioned deployments, Concentric AI integrates with the compliance interfaces the major AI vendors provide.
Claude Compliance API. Anthropic’s Compliance API gives enterprise teams programmatic access to Claude Enterprise conversation content, covering chats, uploaded files, and project contents, alongside an activity event feed. Concentric AI logs every prompt, response, and attachment across both the browser and desktop versions of Claude, including historical activity. Semantic classification then applies to that content, so security teams can identify sensitive data inside Claude interactions and act on it under the same policy framework governing everything else. Anthropic lists the integration among its supported compliance partners.
Microsoft Copilot. Copilot inherits whatever permissions already exist across Microsoft 365. A user who can technically reach a sensitive file through overprovisioned access will find Copilot happy to summarize it for them. Concentric AI’s Data Risk Report found that 16% of business-critical data is overshared, averaging roughly 802,000 files at risk per organization.
Concentric AI maps sensitive data across SharePoint, OneDrive, Teams, and Exchange, finding inappropriate access and sharing, and remediating it before Copilot puts that content in front of the wrong person. Monitoring extends to Copilot prompts and the data access requests made to service them. For organizations already running Microsoft Purview, the two work together, with Purview enforcing policy against classifications Concentric AI has verified.
ChatGPT Enterprise Compliance API. OpenAI’s Compliance Platform exposes workspace logs and conversation content, including prompts, responses, uploaded files, GPT configurations, and workspace metadata. OpenAI swapped out its conversation logs system in March 2026 and retired the old stateful route in June, so integrations built against the previous endpoint need to be migrated. Concentric AI ingests the content and classifies it semantically, so you can see what sensitive data moved through the workspace and when.
These AI tools draw on data that lives somewhere else first, which pulls those source systems into your AIMS scope too. Integrations can span Google Workspace, Salesforce, Slack, Snowflake, Box, and every other environment where sensitive data lives before an AI tool ever reaches it.
What this gives you at audit time
With Concentric AI, you end up with a continuous, timestamped record tying data categories to AI systems, users, and enforcement actions. That supports the data governance controls, responsible use controls, impact assessment inputs, and monitoring requirements in clause 9.
Build the evidence layer your AIMS depends on
ISO 42001 asks you to prove control over how AI touches your data. That proof starts with knowing what data you have, where it lives, and which AI tools are consuming it.
Powered by the Semantic Intelligence™ Engine’s patented language models, the platform discovers, governs, and protects business-critical and compliance-sensitive data at rest, in motion, and as it used by AI through a consistent policy framework.
Request a demo to see how Concentric AI supports AI governance and compliance requirements across sanctioned and unsanctioned AI.
Frequently asked questions
How long does ISO 42001 certification take?
Most organizations spend three to nine months building and running the management system before a certification audit, depending on existing ISO maturity. Companies already certified to 27001 move faster, since the governance structure, internal audit function, and management review cadence already exist.
Is ISO 42001 mandatory?
No. It is voluntary. Market pressure is doing the work instead, with certification increasingly appearing as a line item in enterprise procurement and vendor security reviews.
Can a small company get certified?
Yes. The organization defines its own scope, so certification can cover a single product line or business unit rather than the whole company. Resource load remains the main constraint for teams without existing ISO infrastructure.
What is the difference between ISO 42001 and the National Institute of Standards and Technology Artificial Intelligence Risk Management Framework (NIST AI RMF)?
NIST AI RMF is a voluntary framework with no certification path. ISO 42001 is a certifiable standard. Many organizations use both, with the RMF informing risk practice and 42001 providing the auditable structure around it.
Does ISO 42001 cover shadow AI?
By implication, yes. Responsible use controls and scope definition both require knowing which AI systems operate in the organization. Unsanctioned tools sit outside your documented AIMS while continuing to consume company data, which is exactly what an audit is built to catch.