ISO 27001 vs ISO 42001: Do You Need Both?

ISO/IEC 27001 and ISO/IEC 42001 are complementary, not interchangeable. ISO 27001 is the standard for an information security management system (ISMS): it protects the confidentiality, integrity and availability of information. ISO 42001 is the standard for an AI management system (AIMS): it governs how you develop, provide or use AI systems responsibly, including impacts on people outside your organization. If you build or deploy AI agents and already hold ISO 27001, you do not start over. You extend the same management system with AI-specific scope, an AI impact assessment process and AI controls, and you reuse most of your existing evidence.

The short comparison

	ISO/IEC 27001	ISO/IEC 42001
Manages	Information security risk	Risks and impacts of AI systems
Protects	Information assets (CIA triad)	The organization, individuals and society affected by AI
Management system	ISMS	AIMS
Structure	Harmonized structure, clauses 4 to 10	Same harmonized structure, clauses 4 to 10
Annex A	93 information security controls in four themes (2022 edition)	AI-specific controls covering policy, resources, impact assessment, lifecycle, data, information for interested parties, use and third parties
Distinctive requirement	Statement of Applicability against Annex A	AI system impact assessment, plus a Statement of Applicability
Certifiable	Yes	Yes
Typical buyer pressure	Security questionnaires, enterprise procurement	AI-specific due diligence, regulatory alignment such as the EU AI Act

The shared structure is the reason combining them works. Clauses on context, leadership, planning, support, operation, performance evaluation and improvement have the same shape in both. Internal audit, management review, document control and corrective action can run as one process with two scopes.

Where they overlap

A mature ISMS already gives you a large share of what an AIMS needs:

Risk methodology. Your 27001 risk assessment process (likelihood, impact, treatment, risk owner) carries over. You add AI-specific risk sources.
Access control. Least privilege, identity lifecycle and privileged access reviews apply to AI agents as non-human identities.
Supplier management. Model providers, MCP server vendors and AI platform suppliers fit into your existing third-party risk process.
Logging and monitoring. 27001 logging controls produce the raw material 42001 needs for monitoring AI system behavior.
Incident management. Agent misuse, prompt injection and data leakage through an agent are security incidents. The same playbooks apply, with AI-specific triggers added.
Internal audit and management review. One audit programme, one review cycle, with AI added to the agenda.

What 42001 adds that 27001 does not cover

ISO 27001 asks whether information is secure. It does not ask whether an AI system's output is fair, whether people affected by a decision were informed, or whether the training data was fit for purpose. ISO 42001 does. The main additions:

AI system impact assessment. You assess consequences of the AI system for individuals and society, not only for the organization. We cover the method in ISO/IEC 42005 for AI impact assessment.
AI lifecycle controls. Requirements and design, verification and validation, deployment, operation and monitoring of AI systems, documented per system.
Data for AI. Data quality, provenance and preparation as they affect AI behavior, which is a different question from data confidentiality.
Information for interested parties. What users and affected people are told about the AI system and how they can report problems.
Responsible use. Intended use, objectives for responsible use, and human oversight of AI systems in operation.
AI-specific roles. Whether you are an AI developer, provider or user changes which controls apply.

For agents specifically, the human oversight and lifecycle controls become concrete fast: who approves an agent going to production, which actions require a human, and how you know which agent version made a given decision. See ISO 42001 for AI management systems for a clause-by-clause walk-through.

Do you need both?

It depends on what you do with AI and who asks.

Your situation	Recommendation
No production AI, AI only through standard SaaS features	ISO 27001 alone; treat AI vendors under supplier management
ISO 27001 certified, deploying internal AI agents with access to business data	Extend the ISMS with AI inventory, impact assessments and oversight controls; consider 42001 conformance
Building AI products or agents your customers rely on	Both. Customers will ask for 27001 for security and increasingly for 42001 for AI governance
Selling into the EU with AI systems that may be high-risk under the AI Act	Both. 42001 evidence overlaps with AI Act risk management, record-keeping and oversight duties

A note on the EU AI Act: ISO 42001 is not a harmonized standard under the Act, and holding the certificate does not by itself prove conformity. It does produce much of the documentation the Act asks for. The Digital Omnibus, adopted as Regulation (EU) 2026/1744, moved the high-risk application dates: 2 December 2027 for Annex III systems and 2 August 2028 for Annex I products. Plan against those dates.

If you already report on SOC 2, the same reasoning applies. We compare those two in SOC 2 vs ISO 42001 for AI agent programs.

Gap checklist: from ISO 27001 to ISO 42001

Use this list if you hold 27001 and want to extend into an AIMS.

Scope. Define which AI systems and agents are in the AIMS scope, and your role for each (developer, provider, user).
AI policy. Publish an AI policy aligned with your information security policy, signed off by top management.
AI inventory. Maintain a live list of AI systems, agents, models, tools and MCP servers, including ones not formally registered. An AI bill of materials is a practical format.
Impact assessments. Run and record an AI system impact assessment per in-scope system, and repeat it when the system changes.
Risk register. Add AI risk sources (prompt injection, unsafe autonomy, data leakage through outputs, model drift) to the existing register.
Human oversight. Document which agent actions need human approval and keep the approval records.
Lifecycle records. Keep version history and change approvals per AI system.
Monitoring. Log AI inputs, outputs and tool calls with enough detail to reconstruct a decision.
Evidence integrity. Make audit records tamper-evident so auditors can trust them.
Statement of Applicability. Produce a 42001 SoA alongside the 27001 one, with justified inclusions and exclusions.
Internal audit. Extend the audit programme to AIMS clauses and Annex A controls.

Where you sit on this list is a good proxy for your AI governance maturity.

How to run them as one integrated system

Teams that run both well tend to share four habits:

One risk register, with a tag for AI risks instead of a second spreadsheet.
One evidence store. A single log of agent actions serves as 27001 monitoring evidence and 42001 operational evidence.
One audit calendar. Combined internal audits, with separate checklists per standard.
One management review, with an AI section covering impact assessment outcomes, incidents and oversight metrics.

The cost of running both is mostly the AI-specific delta, not a second management system.

Where Praesidia fits

The 27001-to-42001 gap is mostly evidence about AI agents that a classic ISMS does not collect. Praesidia covers that evidence in three steps:

Discover: an inventory of AI agents, including OpenTelemetry-based discovery and shadow MCP server detection, for your AIMS scope and AI inventory.
Secure: guardrails that block, redact or warn, MCP tool scoping, a BYOK gateway, human-in-the-loop approvals and a kill switch, which give you the oversight controls you document.
Prove: a signed, tamper-evident audit log and executive reports that serve as evidence in both audits.

Praesidia does not run your impact assessments or write your AI policy. It produces the operational records those documents point to.

Key takeaways

ISO 27001 and ISO 42001 solve different problems with the same management system skeleton. If you deploy AI agents and already hold 27001, the efficient path is to extend the ISMS: add AI scope, an AI inventory, impact assessments and oversight records, and reuse your risk, audit and review processes. For the broader governance programme, see our AI governance guide.