SEC-02 / Frameworks

AI and OT security framework crosswalk for manufacturers

No single framework covers AI governance and OT security. This crosswalk shows which framework answers which question, how they compare on scope and effort, and a pragmatic stack for EU and US mid-market manufacturers.

8 min read Updated CISOsCIOsOT engineersPlant managers

Key takeaways

  • Laws tell you what you must achieve (NIS2, CRA, EU AI Act). Standards and frameworks tell you how (IEC 62443, ISO 27001, ISO/IEC 42001, NIST).
  • Most control themes, such as asset inventory, risk assessment, supplier management and incident response, repeat across all nine frameworks. Build each once and map it.
  • For OT security, IEC 62443 is the engineering reference; NIST CSF 2.0 or ISO 27001 is the management wrapper.
  • For AI, use the NIST AI RMF for the risk method and ISO/IEC 42001 for the management system; the EU AI Act sets the legal floor in Europe.
  • A typical mid-market plant needs one management system, one OT engineering standard, one AI method and a clear view of which laws apply.

A mid-market manufacturer adding AI to its plants runs into nine names at once: IEC 62443, NIST CSF 2.0, NIST SP 800-82, ISO/IEC 27001, NIS2, the Cyber Resilience Act, the NIST AI RMF, ISO/IEC 42001 and the EU AI Act. They overlap heavily, but each answers a different question. This crosswalk lays them side by side so you can build controls once and map them many times.

Which framework answers which question

Question you are asking Framework that answers it
How do I engineer security into a plant network and its components? IEC 62443
How do I organize and communicate a cybersecurity program? NIST CSF 2.0
What is different about securing OT, in practical detail? NIST SP 800-82 Rev. 3
How do I run a certifiable security management system? ISO/IEC 27001
Am I legally required to manage cyber risk and report incidents in the EU? NIS2
Must the products and machines I sell meet cybersecurity requirements? Cyber Resilience Act (CRA)
How do I identify and manage AI risk in practice? NIST AI RMF
How do I run a certifiable AI management system? ISO/IEC 42001
Is my AI use lawful in the EU, and what duties do I have? EU AI Act

The big comparison table

Status as of November 2026. Effort is a rough planning view for a single-to-few-plant manufacturer with lean IT/OT staff, not a benchmark.

Framework Scope Mandatory? Certifiable? Who it is for Effort for a mid-market plant
IEC 62443 (series) Security of industrial automation and control systems: asset owners, integrators, product suppliers Voluntary, often required by customers or referenced by laws Yes, for products, components and processes via schemes; less common for asset owners OT engineers, integrators, OEMs High if done fully; medium for a focused zones-and-conduits and SL target effort
NIST CSF 2.0 (Feb 2024) Whole-enterprise cybersecurity outcomes in six functions: Govern, Identify, Protect, Detect, Respond, Recover Voluntary No Executives, CISOs Low to medium; good for a profile and roadmap
NIST SP 800-82 Rev. 3 (Sept 2023) OT security guidance, including an OT overlay of security controls Voluntary No OT engineers, security architects Low as reference; medium if used as a control baseline
ISO/IEC 27001:2022 Information security management system Voluntary; often contractually required Yes CISOs, CIOs Medium to high for first certification
NIS2 (Directive (EU) 2022/2555) Cyber risk management, governance and incident reporting for essential and important entities, including many manufacturing sectors Mandatory if in scope, via national law No (compliance, not certification) Boards, CISOs Medium; depends on national law and entity size
Cyber Resilience Act (Regulation (EU) 2024/2847) Cybersecurity of products with digital elements placed on the EU market Mandatory for in-scope products Conformity assessment and CE marking, not organizational certification OEMs, machine and device makers High for product makers; low for pure users
NIST AI RMF 1.0 (Jan 2023), plus GenAI Profile AI 600-1 AI risk management method: Govern, Map, Measure, Manage Voluntary No AI owners, risk, quality Low to medium
ISO/IEC 42001:2023 AI management system Voluntary Yes Quality, CIOs, compliance Medium; lower if ISO 9001/27001 exist
EU AI Act (Regulation (EU) 2024/1689, amended by Reg. (EU) 2026/1744) Rules on AI systems by risk tier; prohibitions, high-risk duties, transparency, GPAI Mandatory in the EU Conformity assessment for high-risk systems, not organizational certification Providers and deployers of AI Low for minimal-risk deployers; high for high-risk providers

Note: Key dates checked in November 2026. NIS2: the transposition deadline was 17 October 2024; as of November 2026, most member states had transposed it but a few, including France, had not finished, so national rules still vary. CRA: reporting obligations for manufacturers have applied since 11 September 2026, and the main requirements apply from 11 December 2027. EU AI Act: high-risk obligations apply from 2 December 2027 (Annex III) and 2 August 2028 (Annex I) after the AI Omnibus. See the NIS2, CRA and EU AI Act pages for detail.

How the layers fit

  LAW (what you must achieve)
    NIS2 . CRA . EU AI Act . Machinery Reg.
  ------------------------------------------------
  MANAGEMENT SYSTEMS (how you run it)
    ISO/IEC 27001   +   ISO/IEC 42001
  ------------------------------------------------
  METHODS / OUTCOMES (how you think about risk)
    NIST CSF 2.0        NIST AI RMF (+ AI 600-1)
  ------------------------------------------------
  ENGINEERING (how you build and configure)
    IEC 62443           NIST SP 800-82r3

Lower layers produce the evidence upper layers ask for. A zones-and-conduits drawing from IEC 62443 is evidence for NIS2 risk measures, ISO 27001 network controls and CSF Protect outcomes all at once.

Control themes mapped across frameworks

This is a general mapping of where each theme is addressed, not a clause-by-clause copy. "Strong" means the framework treats it in depth; "Light" means it is mentioned or implied; "N/A" means it is out of scope.

Control theme IEC 62443 NIST CSF 2.0 SP 800-82r3 ISO 27001 NIS2 CRA NIST AI RMF ISO 42001 EU AI Act
Asset / system inventory Strong (system under consideration) Strong (Identify) Strong Strong Required (asset management in Art. 21) Product component inventory, incl. SBOM Strong (AI inventory under Govern/Map) Strong (AI systems in scope) Implied; deployers must know their systems
Risk assessment Strong (zone risk, SL targets) Strong (Identify, Govern) Strong Strong Required (risk-based measures) Required for product design Core (Map, Measure) Required (AI risk and impact assessment) Required for high-risk providers
Access control Strong Strong (Protect) Strong Strong Required Required in product Light Light Light (security of high-risk systems)
Supplier management Strong (supplier and integrator requirements) Strong (supply chain risk under Govern) Strong Strong Required (supply chain security) Manufacturer due diligence on components Strong (third-party risk) Strong (third-party relationships) Value-chain duties; role shifts
Incident response and reporting Strong Strong (Respond, Recover) Strong Strong Required, with reporting deadlines Required reporting of exploited vulnerabilities and severe incidents Strong (Manage) Strong (nonconformity, communication) Serious incident reporting for high-risk
Monitoring Strong Strong (Detect) Strong Strong Implied Post-market vulnerability handling Strong (Measure) Strong (performance evaluation) Post-market monitoring; deployer monitoring and logs
Human oversight N/A Light Light (safety interplay) N/A N/A N/A Strong Strong Required for high-risk
Data governance Light Light (data security) Light Strong (information) Light Light Strong Strong (data for AI) Required for high-risk training data

How to read it:

  • Security themes (access, monitoring, incident response) are strongest in IEC 62443, CSF, 800-82 and 27001. AI frameworks assume you already have them.
  • AI themes (human oversight, data governance for training data, impact on people) live only in the AI RMF, 42001 and the AI Act.
  • Inventory, risk, suppliers and incidents appear everywhere. Build one asset and AI inventory, one risk method with an AI extension, one supplier process and one incident process, and map them.

Plant-floor reality: Three auditors asking for three asset inventories is how lean teams burn out. Keep one inventory with columns for OT asset, AI system, product with digital elements and NIS2 relevance. Every framework gets a filtered view of the same list. Start from your OT asset inventory.

A pragmatic stack for a mid-market manufacturer

EU variant (100 to 2,500 staff, sells or operates in the EU)

  1. Legal map first. Confirm whether you are in NIS2 scope under your national law, whether you sell products with digital elements (CRA), and how your AI uses classify under the AI Act.
  2. Management system: ISO/IEC 27001 if customers already ask for it; otherwise NIST CSF 2.0 as a lighter structure, with 27001 as a later step.
  3. OT engineering: IEC 62443 for zones and conduits, security level targets and supplier requirements, with SP 800-82r3 as practical reference.
  4. AI: NIST AI RMF for the risk method, an AI use case register, and AI Act classification. Add ISO/IEC 42001 into the same management system when AI becomes material to your product or customers.
  5. Product makers: add CRA and Machinery Regulation conformity work to engineering.

US variant

  1. Program structure: NIST CSF 2.0, with a target profile agreed with leadership.
  2. OT engineering: IEC 62443 plus SP 800-82r3.
  3. Certification only where customers require it: ISO/IEC 27001 for security, ISO/IEC 42001 for AI.
  4. AI: NIST AI RMF and the Generative AI Profile; the EU AI Act still matters if you sell AI-enabled products into the EU or operate EU sites.
Size of team Minimum viable stack
1 to 2 people covering IT and OT CSF 2.0 profile, IEC 62443 zones and conduits for critical lines, AI RMF-based register and triage
Small security team plus OT champion per plant Add ISO/IEC 27001 and a formal IEC 62443 risk assessment
AI material to products or customers Add ISO/IEC 42001 into the 27001 system

Common mistakes

  • Picking one framework to cover everything. Laws, management systems and engineering standards answer different questions.
  • Running parallel programs. Separate NIS2, 27001 and AI projects with separate inventories and risk registers.
  • Treating NIST and ISO as competitors. CSF and the AI RMF give structure; ISO gives certifiable management systems. They stack.
  • Ignoring product law. If you build machines or devices, the CRA and Machinery Regulation are product obligations, not IT policy.
  • Leaving AI out of OT security. AI vision and edge analytics add new conduits. Treat them under securing AI in OT networks.

Where to start

  • List which laws apply to you (NIS2, CRA, AI Act) and write down why.
  • Choose one management structure (CSF 2.0 or ISO/IEC 27001) and one OT engineering reference (IEC 62443).
  • Create one combined inventory for OT assets and AI systems.
  • Extend your risk method with AI using the NIST AI RMF.
  • Map controls once in the framework crosswalk workbook and reuse the evidence for every audit.

Note: This is general information, not legal advice. Confirm which laws apply to your organization with counsel.

Sources and further reading

  1. NIST Cybersecurity Framework 2.0 (NIST)
  2. NIST SP 800-82 Rev. 3, Guide to Operational Technology Security (NIST CSRC)
  3. The CRA Single Reporting Platform is launched (ENISA)
  4. AI Act: regulatory framework and application timeline (European Commission)
  5. NIST AI Risk Management Framework (NIST)

This guide is general information for manufacturers, not legal advice. Confirm regulatory obligations with qualified counsel.

FAQ

Frequently asked questions

Should a manufacturer choose IEC 62443 or ISO 27001?

Usually both, at different layers. ISO/IEC 27001 runs the information security management system across the company and is widely certified. IEC 62443 gives OT-specific engineering guidance such as zones and conduits, security levels and supplier requirements. Many plants certify 27001 at company level and use 62443 to design and assess the plant network.

Does NIS2 require ISO 27001 or IEC 62443 certification?

NIS2 does not require any specific certification. It requires appropriate and proportionate risk management measures and incident reporting, with the detail set in each member state's national law. Using ISO/IEC 27001 and IEC 62443 is a common way to show you meet those measures. Some national laws reference particular standards, so check your country's transposition.

Which AI framework should a mid-sized manufacturer start with?

Start with the NIST AI RMF for the method, because it is free and practical, and an AI use case inventory. If you sell into the EU or use AI on workers, check the EU AI Act classification at the same time. Add ISO/IEC 42001 when you want a certifiable management system, typically after your first high-impact AI use cases are under control.

Are the NIST frameworks relevant for EU manufacturers?

Yes, as free guidance. NIST CSF 2.0 and SP 800-82 are widely used outside the US for structure and OT detail, and the AI RMF is a useful method for AI risk. They do not replace EU legal obligations under NIS2, the CRA or the AI Act, but they help you implement them.

Related guides

Know where your plants stand

Score your AI governance, OT security and data readiness in about 10 minutes each. Results stay in your browser.