AI · Guide
ISO 42001 certification in India
ISO 42001 certification shows that a company governs its AI through an audited AI management system. In India it is voluntary, but enterprise buyers increasingly ask AI vendors for it. A first certification usually takes four to nine months, and less if you already run ISO 27001, because the two share the same management system structure.
Key takeaways
- ISO/IEC 42001:2023 is certifiable by accredited certification bodies; Annex A contains 38 AI-specific controls.
- The work centres on an AI inventory, AI risk assessment and AI system impact assessment.
- If your AI uses personal data, the DPDP Act applies on top of ISO 42001.
- Exporters selling AI into Europe will find ISO 42001 a practical way to organise EU AI Act evidence.
Who needs ISO 42001 certification in India?
Three groups are asking about it most:
- AI and SaaS product companies selling to banks, insurers and global enterprises. Procurement teams want proof that models are governed, not just a slide about responsible AI.
- IT services firms building AI systems for clients, who need a repeatable way to show how they control data, testing and handover.
- Enterprises using AI in decisions about customers or staff, such as credit scoring, claims triage or hiring, where the consequences of a bad model are real.
Nothing in Indian law requires ISO 42001 today. The pull comes from customers and, for exporters, from regulation abroad.
What are the ISO 42001 requirements?
The body of the standard (clauses 4 to 10) is a management system, and will look familiar to anyone who has run ISO 27001: context, leadership, planning, support, operation, performance evaluation and improvement. What is new sits in planning and operation: AI risk assessment, AI risk treatment and the AI system impact assessment.
Annex A then lists 38 controls grouped under nine objectives:
| Annex A area | What it covers in practice |
|---|---|
| Policies related to AI | An AI policy aligned with your other policies, reviewed regularly |
| Internal organisation | Roles, responsibilities and a way to report concerns about AI |
| Resources for AI systems | Data, tools, computing and people documented per system |
| Assessing impacts of AI systems | Impact assessments on individuals, groups and society |
| AI system life cycle | Requirements, design, verification, deployment, monitoring, technical documentation |
| Data for AI systems | Data quality, provenance and preparation |
| Information for interested parties | What users and affected people are told, and how incidents are communicated |
| Use of AI systems | Responsible use and intended-use boundaries |
| Third-party and customer relationships | Suppliers of models and data, and responsibilities shared with customers |
What are the steps to certification, and how long do they take?
| Step | Typical duration | What you produce |
|---|---|---|
| 1. Gap assessment | 2 to 4 weeks | Gap report against clauses and Annex A |
| 2. Scope and AI inventory | 2 to 4 weeks | List of AI systems, owners, purposes, data and suppliers |
| 3. Risk and impact assessments | 4 to 8 weeks | AI risk register, impact assessments, treatment plan, Statement of Applicability |
| 4. Policies and controls | 4 to 8 weeks | AI policy, lifecycle procedures, supplier terms |
| 5. Operate and record | 1 to 3 months | Records that show the system working |
| 6. Internal audit and management review | 2 to 4 weeks | Audit report, review minutes, corrective actions |
| 7. Stage 1 and Stage 2 audits | 4 to 8 weeks | Certificate, valid three years with yearly surveillance |
The single biggest variable is the AI inventory. Companies are often surprised by how many AI systems they use once they count third-party tools with AI features, not just the models they build.
How much of ISO 27001 can you reuse?
A lot. Document control, internal audit, management review, competence and corrective action carry over unchanged. Many Annex A security controls, such as access management, logging, supplier security and change management, support ISO 42001 objectives directly. Certification bodies commonly audit the two together. The genuinely new work is the AI inventory, impact assessments and lifecycle controls.
How does ISO 42001 interact with the DPDP Act?
If an AI system processes personal data, the DPDP Act applies regardless of your ISO status. Purpose limitation matters most: personal data collected for one purpose cannot simply be reused to train a model for another without a valid basis. Significant Data Fiduciaries must also check that algorithmic software does not risk people's rights. A DPIA and an ISO 42001 impact assessment can share much of the same analysis.
What does certification cost?
Three parts: the certification body's audit days (driven by headcount, sites and the number of AI systems in scope), any consultant support, and internal time. Reusing ISO 27001 lowers the internal effort considerably. Ask two or three accredited bodies for quotes, and check that each is accredited for ISO 42001 specifically, not just ISO 27001.
Common pitfalls
- Scoping only the models you build and ignoring AI features inside bought software.
- Writing an AI policy that nobody can link to an actual decision or control.
- Running impact assessments once and never revisiting them after model changes.
To manage the inventory, assessments and controls in one place, see ISO 42001 compliance software.
Frequently asked questions
Is accredited ISO 42001 certification available in India?
Yes. Several certification bodies operating in India offer ISO 42001 audits. Confirm each is accredited for ISO 42001 by an IAF member accreditation body.
Can we certify ISO 42001 without ISO 27001?
Yes. ISO 42001 stands on its own, but most companies find it faster with ISO 27001 already in place because the management system is shared.
Does ISO 42001 help with the EU AI Act?
It gives you a structured way to organise risk management, documentation and oversight evidence, which helps, but it is not a substitute for the EU AI Act's specific obligations.