Govern AI on the plant floor · OT/IT security · Industrial data

AI governance for manufacturing, without slowing the line.

Practical AI governance for manufacturing: guides, free self-assessments and templates for mid-market manufacturers putting AI into plants, covering IEC 62443, NIST CSF 2.0, the EU AI Act, NIS2 and ISO/IEC 42001, explained in the language of operations.

  • Free, no account needed
  • Assessments run in your browser
  • Vendor-neutral, plant-tested language
  • Sources linked on every guide
PURDUE-REF // AI DATA FLOWS OTAI
> Every AI use case adds a data flow across these levels. Govern the flow, not just the model.

Regulatory clock

August 1, 2024EU AI Act enters into forceIn effectOctober 17, 2024NIS2 transposition deadlineIn effectFebruary 2, 2025AI Act prohibitions and AI literacy applyIn effectAugust 2, 2025AI Act general-purpose AI obligations applyIn effectJune 11, 2026CRA rules for conformity assessment bodies applyIn effectJuly 27, 2026AI Act Digital Omnibus (Reg. 2026/1744) in forceIn effectAugust 2, 2026AI Act transparency duties (Art. 50) applyIn effectSeptember 11, 2026CRA vulnerability and incident reporting startsIn effectJanuary 20, 2027EU Machinery Regulation appliesUpcomingDecember 2, 2027AI Act Annex III high-risk rules apply (after Omnibus)UpcomingDecember 11, 2027Cyber Resilience Act fully appliesUpcomingAugust 2, 2028AI Act high-risk rules for AI in regulated products (after Omnibus)Upcoming
Guides
33
frameworks, governance, OT, data
Self-assessments
4
instant results, nothing sent
Free templates
9
Excel and Word, editable
Glossary terms
90
OT, AI and regulation

00 / START HERE

Start from your question

The situations that land on a plant manager’s or CISO’s desk, and where to go first.

Q-01

“Our OEM wants permanent remote access to the new line”

Say yes to remote support, no to an always-on connection. Route the OEM through a brokered jump host in the IDMZ with named accounts, MFA, per-session approval and recording, and write it into the contract before FAT.

Q-02

“The board just asked for our AI plan”

Start with an inventory of where AI is already used, a simple policy and a named owner, then brief the board on exposure and decisions needed. A credible 90-day plan beats a big strategy deck.

Q-03

“A customer sent us a NIS2 letter asking about our cybersecurity”

First check whether NIS2 applies to you directly or only through your customer's supply chain duties. Either way, answer with evidence mapped to one framework rather than a fresh promise each time.

Q-04

“Our engineers are pasting PLC code into ChatGPT”

Do not just block it. Give engineers an approved tool with proper data terms, set a clear rule on what code, recipes and drawings can be shared, and require review before any AI-written logic goes near a controller.

Q-05

“A vendor is pitching us predictive maintenance”

Check that you have the data first: sensor coverage, failure history in the CMMS and a way to turn alerts into work orders. Then ask how their platform connects to OT, who owns the data and what happens when the model is wrong.

Q-06

“IT wants to connect the historian to the cloud”

It can be done safely if data only flows outward through the IDMZ, ideally from a historian replica, with no inbound path to the control network. Document the flow as a conduit and put it through management of change.

Q-07

“Quality wants AI vision to make release decisions”

Treat it as a high-risk use case until shown otherwise. Validate it like any inspection system, keep a qualified person able to override, monitor for drift and decide up front what triggers a fallback to manual inspection.

Q-08

“A customer asked whether our product complies with the Cyber Resilience Act”

If you sell connected machines, controllers or software in the EU, you are likely a manufacturer under the CRA. Confirm scope, map your SBOM, vulnerability handling and update support to its requirements, and give the customer a dated, honest status.

01 / THE LIBRARY

Four pillars, one governed plant

Governance decides what AI may do. Frameworks give you the shared language. OT security keeps the new data flows safe. Data foundations make the AI worth governing.

02 / MEASURE

Free self-assessments with a 30/60/90-day plan

Concrete, observable answers. Instant scores by dimension, the gaps that matter and the next actions. Your answers never leave your browser.

03 / FRAMEWORKS

Which framework answers which question?

One language for the board, one engineering standard for the plant floor, and a clear view of which laws actually bind you. The crosswalk maps them side by side.

FrameworkCoversStatus
IEC 62443OT securityEngineering standard
NIST CSF 2.0IT + OT securityVoluntary framework
NIST SP 800-82r3OT securityGuidance
ISO/IEC 27001IT securityCertifiable
NIS2Cybersecurity law (EU)Mandatory if in scope
Cyber Resilience ActProduct security law (EU)Mandatory for products
NIST AI RMFAI riskVoluntary framework
ISO/IEC 42001AI management systemCertifiable
EU AI ActAI law (EU)Mandatory, risk-based

04 / DATA FOUNDATIONS

Most industrial AI failures are data failures

A predictive maintenance model is only as good as the failure history in your CMMS. A vision model is only as good as its labels. Moving historian data to the cloud opens a new path out of the plant. That is why data readiness sits inside governance here: ownership, lineage, quality and access are the controls that make AI safe and worth running.

05 / TEMPLATES

Free templates you can use on Monday

Editable Excel and Word files with worked examples. No sign-up.

All templates

Pro Kit

Plant AI Governance Pro Kit

Six working files for mid-market manufacturers that need to govern AI on the shop floor and secure the OT it runs on. Every file is pre-filled with a worked example for a fictional 3-plant, 650-person parts maker, so you can see what good looks like before you replace it with your own data.

See what is inside · €249

06 / WHO IT IS FOR

Built for manufacturers with 1 to 15 plants and lean teams

Plant managers

Know which AI on your site matters, who owns it and what to do when it is wrong, without a compliance department.

OT and controls engineers

Zones, conduits, remote access and edge AI patterns that respect safety, uptime and maintenance windows.

CISOs and CIOs

One governance model across IT and OT, mapped to NIS2, the AI Act and ISO/IEC 42001, sized for a lean team.

Quality, EHS and operations leaders

Rules for AI that touches release decisions, worker data or safety functions, tied to MOC and validation.

FAQ

Frequently asked questions

What is MFG AI Governance?

An independent resource on AI governance, OT and IT cybersecurity and data foundations for manufacturers. It offers plain-language guides to frameworks such as IEC 62443, NIST CSF 2.0, NIS2 and the EU AI Act, free self-assessments and editable templates you can put to work on Monday morning.

Who is this site for?

Mid-market manufacturers, roughly 100 to 2,500 employees across one to fifteen plants, with lean IT and OT teams. It is written for plant managers, controls and OT engineers, CIOs and CISOs, and quality, EHS and operations leaders who need practical answers rather than enterprise-scale programs.

Is it free?

Yes. The guides, assessments and downloadable toolkits are free to use. There is no paywall and no sign-up needed to download the templates.

Does this replace legal advice?

No. The content explains frameworks and regulations to help you plan and ask better questions, but it is not legal advice. Regulatory scope and deadlines depend on your facts and can change, so confirm obligations under the EU AI Act, NIS2, the Cyber Resilience Act and similar laws with qualified counsel.

Which framework should a mid-market manufacturer start with?

For cybersecurity, many start with NIST CSF 2.0 to organize the overall program and use IEC 62443 for the plant floor, especially zones and conduits and supplier requirements. For AI, start with an inventory and a policy, then use NIST AI RMF or ISO/IEC 42001 for structure. Let customer, regulatory and insurer requirements set the priority.

Are my assessment answers stored?

No. The assessments run entirely in your browser. Your answers and scores are computed on your device and are not sent to us or stored on a server. If you want to keep your results, save or print them yourself before you close the page.

Can I use the templates in my company or with clients?

Yes. You can adapt, rebrand and use the templates commercially, including in your own company or in client work. Review and adapt them to your situation before use, since they are starting points rather than finished compliance documents.

Ten minutes to a clear picture of your plants

Start with the AI governance readiness assessment. You will get a score by dimension and a 30/60/90-day plan you can take to your next ops review.