SEC-03 / OT Security

OT asset inventory: the foundation for OT security and AI governance

An OT asset inventory is a maintained list of every device, system and connection in your control environment, with the attributes you need to make security decisions. Without it, segmentation, patching, incident response and AI governance are guesswork.

6 min read Updated OT engineersControls engineersCISOsPlant managers

Key takeaways

  • CISA and partners published 'Foundations for OT Cybersecurity: Asset Inventory Guidance for Owners and Operators' in August 2025. It is a good template for scope, attributes and lifecycle.
  • Combine physical walk-downs, passive network monitoring and carefully controlled active queries. Never run IT-style scans against live controllers.
  • Capture a focused set of fields first: identity, location, function, criticality, versions, protocols, owner and remote access path.
  • Vendor-owned and OEM-supplied assets belong in your inventory, with the vendor named as maintainer.
  • Tag AI-enabled devices and data flows so the same list feeds your AI use-case register.

Ask a plant how many PLCs it has and you will usually get a number. Ask which firmware they run, which ones the OEM can reach remotely, and which feed the new AI quality model, and the room goes quiet. An OT asset inventory closes that gap. It is the least glamorous part of OT security and the one everything else depends on.

What an OT asset inventory is

An OT asset inventory is a maintained record of the hardware, software and connections in your control environment, with the attributes you need to make decisions. It is not a one-off spreadsheet from a consultant's visit. It is a living dataset with an owner, a review cycle and links to change management.

In August 2025, CISA published Foundations for OT Cybersecurity: Asset Inventory Guidance for Owners and Operators with the EPA, NSA, FBI and cyber agencies from Australia, Canada, Germany, the Netherlands and New Zealand. It describes a five-step process:

  1. Define scope and objectives, including governance, roles and boundaries.
  2. Identify assets and collect attributes, through physical inspection and a logical survey.
  3. Create a taxonomy to categorize assets, for example by function or criticality.
  4. Identify additional data sources and store asset data in a centralized, secured system.
  5. Manage the lifecycle, from acquisition through decommissioning.

The guidance's sector examples cover oil and gas, electricity and water, but the method applies directly to manufacturing. It also recommends organizing the taxonomy into zones and conduits in line with ISA/IEC 62443, which links straight to your zones and conduits work.

Does this apply to you?

If you have any control system on a network, yes. You will need an inventory for nearly every framework you might follow. NIST CSF 2.0 starts asset management under Identify (see NIST CSF 2.0). IEC 62443 risk assessment starts by identifying the system under consideration. NIS2 risk-management measures assume you know what you are protecting.

How to discover assets

There are three sources. Use all of them.

Method How it works Strengths Risks and limits
Physical walk-down Engineers open panels, read nameplates, photograph labels Finds unnetworked and serial devices, confirms location Labor-intensive; goes stale
Passive monitoring Sensor on a SPAN port or tap decodes industrial protocols No traffic sent; continuous; sees communication paths Misses silent devices; needs SPAN coverage on the right switches
Active querying Tool sends protocol-native or standard queries to devices Gets firmware, modules, serials, configuration Can upset fragile or old devices if done carelessly
Configuration files Export from PLC projects, SCADA tag databases, switch configs Accurate, rich detail Only as current as the last export

Watch out: Do not point an IT vulnerability scanner at a live control network. Aggressive scans have crashed older PLCs and network modules. If you use active discovery, use OT-aware tools, test on a bench or idle line first, and schedule it with operations.

Most mid-market plants get the best return from a passive OT monitoring tool at a few core switches, combined with a walk-down of each line and an export from engineering projects. Add targeted active queries later for firmware detail.

Which fields to capture

CISA's guidance lists high-priority attributes: communication protocols, asset criticality, asset number, role or type, hostname, IP address, logging, MAC address, manufacturer, model, operating system, physical location, ports and services, and user accounts. Here is a practical starting set for a manufacturer, grouped by purpose.

Group Fields Why
Identity Asset ID/tag, hostname, manufacturer, model, serial Find it and match it to advisories
Location and function Site, line/cell, panel, Purdue level, zone, role (PLC, HMI, drive, edge, SIS) Segmentation and incident scoping
Network IP, MAC, VLAN, protocols, open ports and services Firewall rules and anomaly baselines
Software OS, firmware version, key application versions Patch and vulnerability triage
Risk Criticality (process, safety, quality), safety-related flag Prioritization
Ownership Business owner, technical owner, maintainer (in-house or OEM), contract reference Who decides and who fixes
Access Remote access path, local accounts, logging destination Remote access control and forensics
AI and data Contains AI/ML function (Y/N), sends data externally (Y/N), destination, feeds which models AI governance and data lineage
Lifecycle Install date, end-of-support date, backup location of program/config Replacement planning and recovery

Start with the fields you can keep current. A smaller accurate inventory is worth more than a large stale one.

Vendor-owned and OEM-supplied assets

Mid-market plants run on OEM equipment. A packaging line may arrive as a skid with its own PLC, HMI, industrial PC, switch and LTE router. Contracts often say the OEM maintains it. That does not make it someone else's risk.

  • Inventory it anyway. Record what you can see passively and on the nameplate. Mark the OEM as maintainer.
  • Record the remote access path. Many OEM skids ship with their own remote access router. These belong on your secure remote access list, and usually need to be replaced or brokered.
  • Make data a purchase requirement. New equipment contracts should require an asset list with models, firmware, open ports, accounts, any AI functionality and an SBOM where available. Our OT/AI vendor security questionnaire includes these questions.
  • Note the gaps. If an OEM refuses to share details, record that as a risk with a named owner.

Plant-floor reality: Black-box OEM cabinets are where you will find the surprises: an old Windows PC behind the HMI, an unmanaged switch, a cellular modem nobody knew about. Walk them down with the maintenance tech who services that line.

Using the inventory for AI governance

AI on the shop floor often arrives inside equipment you already bought or are about to buy. A vision system with a trained reject model, a drive with "AI-based" predictive maintenance, an energy optimizer that talks to a cloud service. Your OT inventory is the best place to find them.

Add three fields: contains AI/ML function, sends data to external services, and which models the asset feeds. Then:

  • Export flagged assets into your AI use-case inventory and the AI use-case register.
  • Use the "feeds which models" field during incidents. If a sensor was tampered with or drifted, you know which models to review.
  • Use the external data flag to check that no AI device has an unapproved path out of OT.

Keeping it alive

An inventory decays from the day you finish it. Keep it current by:

  • Tying updates to management of change. No device goes on or comes off the network without an inventory update.
  • Reconciling passive monitoring results against the inventory monthly. New devices are either approved changes or findings.
  • Reviewing criticality and ownership annually with operations.
  • Storing it somewhere secure. An inventory is a map for attackers. Restrict access and keep an offline copy for incident response.

What good looks like

  • Every networked OT asset at each site is recorded with the core fields above.
  • Unnetworked but critical assets (safety controllers, serial devices) are recorded from walk-downs.
  • OEM assets carry maintainer and remote access fields.
  • AI-enabled assets are flagged and linked to the AI register.
  • Monitoring alerts on unknown devices, and the inventory is updated through change management.

Effort and cost

For a single mid-size plant, expect two to six weeks of combined engineering time for the first pass: walk-downs, sensor deployment, data cleanup and owner interviews. A passive monitoring platform is a recurring license, usually priced by site or sensor. Spreadsheets work for a small site. Beyond a few hundred assets or multiple sites, a dedicated tool or CMMS integration pays back in maintenance effort.

Common mistakes

  • Running an IT network scan on the control network.
  • Treating the consultant's report as the inventory.
  • Leaving OEM skids off the list "because they are the vendor's".
  • Capturing 60 fields and maintaining none of them.
  • No link between the inventory and change management.

Monday-morning checklist

  1. Name an owner for the OT asset inventory at each site. See IT/OT convergence governance for where that role sits.
  2. Read the CISA asset inventory guidance and agree your scope and field list.
  3. Export what you already have: PLC projects, SCADA tag lists, switch ARP tables.
  4. Walk down one line end to end, including OEM cabinets, and reconcile.
  5. Plan a passive monitoring pilot at the core OT switch.
  6. Add AI and external data flow fields and flag the first candidates for the AI register.

Sources and further reading

  1. Foundations for OT Cybersecurity: Asset Inventory Guidance for Owners and Operators (CISA and partners, Aug 2025)
  2. Joint guide PDF: Foundations for OT Cybersecurity, Asset Inventory Guidance
  3. NIST SP 800-82 Rev. 3, Guide to Operational Technology (OT) Security
  4. Secure connectivity principles for operational technology (ASD's ACSC, NCSC-UK and partners, Jan 2026)

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

FAQ

Frequently asked questions

What is the difference between passive and active OT asset discovery?

Passive discovery listens to network traffic from a switch SPAN port or tap and identifies devices from what they say, without sending anything. Active discovery sends queries to devices, often using native industrial protocols, to read model, firmware and configuration. Passive is safer for fragile devices but misses quiet ones. Active gives richer data but must be tested and scheduled. Most plants combine both, plus a physical walk-down.

What fields should an OT asset inventory include?

CISA's August 2025 guidance lists high-priority attributes including communication protocols, asset criticality, asset number, role or type, hostname, IP and MAC address, logging, manufacturer, model, operating system, physical location, ports and services, and user accounts. Add firmware version, owner, maintainer, zone, and remote access path. Start with what you can maintain.

Do I need to inventory equipment owned or maintained by an OEM?

Yes. If it is on your network or in your plant, it is in your risk. Record OEM-supplied PLCs, HMIs, industrial PCs, edge boxes and their remote access paths, with the OEM named as maintainer and a contract reference. If the OEM will not share details, record what you can observe passively and note the gap. Make inventory data a contract requirement for new equipment.

How does an OT asset inventory help with AI governance?

Many AI systems on the shop floor arrive inside equipment: vision systems, predictive maintenance modules and smart drives. If your inventory has a field for AI or ML functionality and data flows to external services, it becomes the discovery tool for your AI use-case register. It also tells you which assets feed AI models, which matters for data quality and incident response.

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.