On September 22, 2026, Aurigo Software announced it had achieved ISO/IEC 42001 certification for its artificial intelligence management system. In the announcement, its CISO described the standard as a disciplined way to govern risks such as data leakage, model drift, bias and unreliable outputs, and the company framed the certificate as an independently validated benchmark for customers evaluating AI governance in vendor systems during procurement and due diligence. It is one of a steady run of similar announcements from software and services providers this year.
The trend is welcome. Until recently, buyers asking a vendor how it governs AI received a slide deck. A certificate issued after an independent audit is a real step up. But the way it is being used in procurement — as a binary pass/fail on a questionnaire row — invites the same mistake enterprises made with SOC 2 reports and ISO 27001 badges for years: reading the headline and skipping the scope. Crest.Digital's view is that ISO 42001 deserves to be read the way an auditor reads a SOC report, and then backed by monitoring that runs between audits.
That matters beyond vendors. Organizations pursuing the certificate themselves face the same trap in reverse: a clean audit is not the same as an operating AI governance program. Our earlier work on why an AI inventory has to come before AI governance and why the question is which vendors, not which framework makes the underlying point. A standard organizes the work; it does not do it.
Crest.Digital's TPRM platform stores governance certificates against each vendor, flags scope gaps and expiries, and monitors vendors after onboarding so a certificate is a living data point, not a one-time tick.
Explore the TPRM PlatformWhy the Certification Wave Is Arriving Now
ISO/IEC 42001, published in December 2023, is the first international management-system standard for AI. Its structure will be familiar to anyone who has worked with ISO 27001: leadership commitment, risk and impact assessment, documented controls, internal audit, management review and continual improvement. What it adds is AI-specific content — impact assessment on individuals and society, data and model lifecycle controls, transparency, and oversight of AI suppliers.
Two developments explain the current acceleration. First, certification capacity has matured: the first ANAB-accredited ISO 42001 certifier was announced in September 2024, and further accredited bodies have followed since. Second, ISO/IEC 42006:2025, published in July 2025, set requirements for the bodies that audit and certify AI management systems, including collective competence across AI technology, governance and legal context, and a structured method for calculating audit time that weighs the number of AI systems, high-risk uses and third-party relationships. As summarized by Bright Defense, the standard treats the calculated audit time as a minimum and requires documented justification for deviations. In plain terms, the market is building the equivalent of the assurance infrastructure that surrounds financial audits and information-security certification.
What an ISO 42001 Certificate Proves — and What It Doesn't
A certificate proves that an independent certification body audited a defined management system, within a defined scope, and found it conforming to the standard at the time of the audit, with surveillance audits to follow. That is valuable. It tells a buyer that policy, ownership, risk assessment and internal audit exist for the AI systems in scope, and that someone outside the company has looked.
It does not prove that every AI feature the vendor ships is covered; that a particular model is accurate or unbiased; that the system you are buying behaves the same way after next quarter's model update; or that the vendor's own AI suppliers — the foundation-model providers and data sources underneath — are governed to the same standard. That last point is where third-party risk teams should pay attention. We covered the structural issue in the fourth-party AI model you never contracted with: a vendor's certificate speaks to the vendor's management system, while the model doing the work may belong to someone else entirely.
The same logic applies to your own certificate. The COSO Internal Control framework distinguishes designing controls from monitoring that they keep operating, and the IIA's Global Internal Audit Standards build assurance around independent, evidence-based testing. A management system that passed an audit in March can be quietly out of date by June, especially when AI systems change faster than annual cycles.
Six Checks Before You Rely on a Vendor's ISO 42001 Certificate
These six checks turn a badge into due-diligence evidence. They take a few minutes per vendor once the certificate is in hand, and they are far cheaper than discovering a scope gap after an incident.
Scope: Is Your Product Inside It?
Read the scope statement line by line. Certificates often cover a named platform, business unit or set of sites. If the AI capability you are buying is a newer product line, confirm it is explicitly included.
Certifier: Who Audited, and Under What Accreditation?
Confirm the certification body is accredited for ISO 42001 and note whether it operates to the ISO/IEC 42006 requirements. Depth of audit varies, and the certifier is a proxy for that depth.
Dates: Valid, Surveilled, Current?
Check issue date, expiry and surveillance-audit status. A certificate in its second year with a missed surveillance audit is a different signal from one freshly issued.
Statement of Applicability: What Was Excluded?
Ask which Annex A controls were applied and which were excluded, with the vendor's justification. Exclusions around third-party AI, data provenance or human oversight deserve follow-up questions.
Supply Chain: Whose Models Sit Underneath?
Identify the foundation models, data sources and sub-processors behind the AI features, and ask how the vendor governs them. The certificate rarely answers this on its face.
Change Control: What Happens Between Audits?
Ask how model updates, prompt changes and new data sources are assessed before release, and what triggers notification to customers. Contractual change-notification rights matter more than the badge here.
Where an AI-governance certificate does not exist, the same six questions still work as a structure for evidence requests; our guidance on AI governance in vendor due diligence goes deeper on the questionnaire side, and why the questionnaire is the least interesting part of TPRM explains why documentary answers should be corroborated rather than collected.
Crest.Digital's agents monitor the controls and vendor signals that change between audits — inventory completeness, approvals, access, supplier changes and certificate status — and route exceptions to named owners with the evidence retained.
The Gap Between Audits: Where AI Governance Actually Fails
Management-system certification works on a cycle: an initial audit, then surveillance and recertification at intervals. AI systems do not follow that cycle. A model provider upgrades a model; a team connects a new tool to an agent; a data source changes; an owner leaves; an impact assessment completed at launch is never revisited. None of these events is visible to an annual audit, and each can silently move a system outside the assumptions on which its governance was approved.
This is the same structural gap internal audit has long managed with continuous controls monitoring for finance and access controls. Applied to an AI management system, it means testing a defined set of controls on an ongoing basis rather than sampling them once a year. Four examples carry most of the value: inventory completeness (new AI use discovered versus registered, the shadow-AI problem we discussed in shadow AI as the new shadow vendor problem), assessment currency (impact and risk assessments reviewed after material change), access and change exceptions (who can modify prompts, models and connected tools, as in our piece on AI agent identity and access governance), and evidence integrity (logs and approvals complete and retained, per our guidance on audit trails for AI agents).
This is also where independent verification enters. A monitoring layer that is separate from the teams running the AI gives internal audit evidence it can rely on, the argument we made in why AI needs its own independent assurance layer. The auditor stays the decision-maker; the agents do the repetitive testing and follow-up that no audit team can sustain by hand.
From Certificate to Continuous Governance: A Six-Step Playbook
Crest.Digital delivers this as a configurable “Risk Automation Pod” rather than a bespoke build. The reusable stack — integration connectors, agent orchestration, a workflow and approval engine, a rules and risk-scoring engine, an evidence repository, AI logs and explainability, and role-based access — is common across engagements; the client-specific work is the data mapping, control logic, thresholds and reporting. The result works whether the starting point is your own ISO 42001 program, an ISO 27001 base you are extending, or a portfolio of vendors presenting certificates of varying quality.
The Discover → Design → Connect → Deploy → Validate → Transfer Model
- Discover: Inventory AI systems, owners and suppliers, and map them to the scope of any certification you hold or rely on.
- Design: Define the AIMS controls to test continuously, the evidence each needs, thresholds, owners and human review points.
- Connect: Integrate GRC, identity, ticketing, document and procurement systems so tests read systems of record, not self-reports.
- Deploy: Start with the highest-consequence AI use cases and the vendors whose AI sits inside critical processes.
- Validate: Run parallel testing against known-good and known-bad cases to set accuracy thresholds before results are relied on.
- Transfer or manage: Hand the monitoring framework to internal audit or risk, or continue it as a Crest.Digital-managed service.
For organizations already running information-security management, the practical shortcut is reuse: the document control, internal audit, management review and corrective-action machinery built for ISO 27001 carries over to ISO 42001, so continuous monitoring can target the AI-specific controls rather than rebuild the whole system. For India-headquartered groups and global capability centers serving regulated clients, the same evidence set also supports customer audits and the regulator-facing expectations that increasingly accompany AI use in financial services, in line with the direction set out in our analysis of building operational AI governance before regulation forces it.
Standards Are Not Law: Keep the Two Maps Separate
ISO 42001 is voluntary; the EU AI Act is binding, with risk-tiered duties for providers and deployers. A certificate can support parts of AI Act readiness — documented risk management, oversight and record-keeping overlap with the standard — but it is not a conformity assessment under the Act and does not demonstrate legal compliance by itself. We tracked how timelines have shifted in the AI Act deadline moved, the compliance problem didn't. The sound approach is two maps: one of the standard's controls, one of legal obligations, with a shared evidence layer so that one test result can serve both.
For boards and audit committees, the takeaway is a short set of questions. Which of our vendors hold AI-governance certificates, and for which products? Do we hold one, and what does its scope exclude? And between audits, what tells us, with evidence, that the controls still operate? The ISACA guidance ecosystem and COBIT-style control mapping give a common language for these questions; what most organizations lack is the continuous evidence to answer them. A certificate is where credible AI governance begins, not where it ends.
Crest Perspective
AI governance programmes that treat AI risk as a technology problem will be underprepared for the regulatory environment taking shape in 2025–26. The EU AI Act, NIST AI RMF, and emerging FCA guidance all frame AI risk as a governance and accountability question first — the technology controls follow from having clear answers to who is responsible, what decisions the AI influences, and what the escalation path is when the system behaves unexpectedly.
Frequently Asked Questions
ISO/IEC 42001:2023 is the first international management-system standard for artificial intelligence. It requires an organization to establish, implement, maintain and continually improve an AI management system (AIMS) covering policy, roles, risk and impact assessment, lifecycle controls, supplier oversight and internal audit. Certification proves that an accredited or recognized certification body audited a defined scope and found the management system conforming at that point in time. It does not prove that every AI system the company runs is safe, unbiased or accurate, and it does not cover AI features or products that sit outside the stated scope. The certificate is evidence of a governance process, not a guarantee of any individual model's behavior.
Five things. First, the scope statement: which products, AI systems, business units and sites are included, and whether the product you actually buy is inside it. Second, the certification body and whether it is accredited for ISO 42001, noting that ISO/IEC 42006:2025 now sets competence and audit-time requirements for bodies that certify AI management systems. Third, the validity dates and surveillance-audit status. Fourth, the Statement of Applicability, or at least a summary of which Annex A controls were applied and which were excluded, and why. Fifth, the vendor's open nonconformities and how its management system handles AI changes between audits. Request the certificate and scope in writing and file them against the vendor record rather than relying on a website badge.
ISO 27001 is a management-system standard for information security; ISO 42001 applies the same management-system logic to AI-specific risks such as bias, transparency, model drift, data quality and impact on individuals, and the two are designed to integrate, so organizations with ISO 27001 can often reuse structure and processes. The EU AI Act is binding law with risk-tiered obligations for providers and deployers of AI systems, whereas ISO 42001 is a voluntary standard. A certificate can help evidence parts of AI Act readiness, but it is not a conformity assessment under the Act and does not by itself demonstrate compliance with any legal obligation. Treat the standard as a governance backbone and the regulation as a separate obligation map.
Usually not. Certification involves an initial audit and periodic surveillance and recertification audits, so evidence of the management system is sampled at intervals. AI systems can change far more often than that: a model update, a new data source, a changed prompt or a newly connected tool can alter behavior without triggering any audit. Between audits, organizations need their own continuous monitoring of the controls that matter, such as inventory completeness, impact-assessment currency, access to AI systems, supplier changes and incident logging, with exceptions routed to owners and evidence retained. This is the gap that continuous controls monitoring is designed to close, and it is what turns a certificate from a snapshot into an operating discipline.
Crest.Digital's Agentic Risk & Continuous Assurance practice builds customized agents that connect to your existing GRC, identity, ticketing, document and procurement systems and test the controls behind an AI management system on an ongoing basis: AI inventory completeness, owner assignment, impact-assessment and approval currency, access and change exceptions, and evidence collection mapped to the clauses and controls you are working to. On the vendor side, the TPRM platform captures AI-governance certificates and scope against each vendor record, flags expiry and scope gaps, and monitors for changes after onboarding. Delivery follows a configurable Risk Automation Pod model, and the work can be handed to internal audit or run as a managed service. Crest.Digital is not a certification body and does not issue or replace ISO 42001 certification.