# Licensing PLC, SCADA and Historian Data for AI Training

> Can a plant license PLC code, SCADA alarm logs and historian data for AI training? What buyers ask for, who owns integrator code and what to remove.

Canonical: https://origindatapartners.com/guides/license-plc-scada-historian-data-ai
By: Origin Data Partners
Status: Published
Published: 2026-11-15
Updated: 2026-11-15
Sources checked: 2026-11-15

## Quick answer

Yes, when the plant owns the records and no customer contract restricts them. Buyers ask for alarm logs and process data with measured results, kept for several years. Treat control code, tag lists, alarm history and time series as four separate assets, each with its own ownership question. Raw sensor values alone are worth little. The value is the link from an event to its cause and its fix. Check integrator contracts and export controls before a sample leaves.

## Key points

- Control code, tags, alarms and time series are four assets with four owners to check.
- A stop linked to its cause and fix is worth far more than the trend alone.
- Integrator-written PLC code belongs to the integrator unless a signed writing says otherwise.
- Setpoints and recipes are trade secrets. Decide what stays out before the export.

## Can a plant license its PLC, SCADA and historian data?

Yes, if the plant owns the records and no customer or vendor contract restricts them. The data has to be linked to what happened on the line, and the plant has to decide first which setpoints and recipes stay private.

A public listing for manufacturing and industrial companies (one buyer's public listing, read 15 November 2026) describes the fit as "plants, machine shops and industrial operators that keep detailed production and maintenance records: CAM programs, work orders, inspection reports, alarm logs and process data with measured results." One of its three qualifying record sets is "machine alarm logs and process data". It asks for "Several years of history" and lists MES and a plant historian product among the systems it expects.

It also says what does not fit: "Data restricted by customer contracts". The terms themselves have plain definitions. A programmable logic controller is "A solid-state control system that has a user-programmable memory for storing instructions" ([NIST glossary](https://csrc.nist.gov/glossary/term/programmable_logic_controller)). SCADA is "A generic name for a computerized system that is capable of gathering and processing data and applying operational controls over long distances" ([NIST glossary](https://csrc.nist.gov/glossary/term/supervisory_control_and_data_acquisition)). A data historian is "A centralized database supporting data analysis using statistical process control techniques" ([NIST glossary](https://csrc.nist.gov/glossary/term/data_historian)).


## Code, tags, alarms and time series: four separate assets

Plants tend to talk about "the SCADA data" as one thing. For licensing it is four, and each has a different author and a different risk. The code is written in the languages of [IEC 61131-3](https://webstore.iec.ch/en/publication/4552), which "consists of two textual languages, Instruction List (IL) and Structured Text (ST), and two graphical languages, Ladder Diagram (LD) and Function Block Diagram (FBD)."

| Asset | What it holds | Ownership question | Main risk |
| --- | --- | --- | --- |
| Control code | PLC projects, ladder logic, function blocks, HMI screens, version history | Who wrote it: staff, an integrator or the machine builder? | Vendor libraries and integrator copyright |
| Tag database | Tag names, descriptions, units, ranges, alarm limits | Usually the plant, built on a vendor or integrator template | Names that reveal customers, products and sites |
| Alarm and event history | Alarm, acknowledge and return times, operator actions, interlock trips | The plant | Operator names and badge numbers |
| Time series | Sensor values by tag and time, batch and lot context | The plant, unless the process is run for a customer | Setpoints and recipes that are trade secrets |

A buyer can use any one of the four. The set is worth most when the alarm history and time series can be joined to the code version that was running that day.


## Is raw sensor data worth anything on its own?

Very little. A year of temperatures with no context is a list of numbers. Value comes from the link between an event, its cause and what was done about it, which is why the listing asks for process data "with measured results".

A model learns to diagnose a line the way a new engineer does: by seeing the trend before a trip, the alarm sequence, the work order that followed and whether the fix held. Each link you can keep raises the value of the set.

| Raises value | Lowers value |
| --- | --- |
| Alarms joined to the downtime reason code and the work order | Alarm floods with no acknowledgement or cause recorded |
| Tag descriptions, units and ranges kept with the values | Bare tag names that only the controls engineer can read |
| Batch or lot IDs that tie a run to its quality result | Time series with no product or batch context |
| Code version history showing what changed after a fault | Only the current program, with no history |
| Raw values at the logged rate, with compression settings stated | Hourly averages exported from a report |
| Several years, covering start-ups, changeovers and upsets | A few clean weeks of steady running |

Downtime logs and work orders are covered in the [industrial maintenance records guide](https://origindatapartners.com/guides/license-industrial-maintenance-records-ai). Inspection and SPC records are covered in the [manufacturing quality records guide](https://origindatapartners.com/guides/license-manufacturing-quality-records-ai).


## Who owns PLC code written by a system integrator?

The integrator, unless a signed writing says otherwise. Paying for the project gives the plant a right to run the code. It does not by itself hand over the copyright, so read the integration contract before you count the code as yours.

The copyright statute explains why. A work made for hire is "a work prepared by an employee within the scope of his or her employment", or a commissioned work in one of nine listed categories "if the parties expressly agree in a written instrument signed by them" ([17 U.S.C. 101](https://www.govinfo.gov/link/uscode/17/101?type=usc&year=mostrecent&link-type=html)). Control software is not on that list. Code your own controls engineers wrote is the company's. Code an outside integrator wrote starts with the integrator.

It can be assigned, but "A transfer of copyright ownership, other than by operation of law, is not valid unless an instrument of conveyance, or a note or memorandum of the transfer, is in writing and signed by the owner of the rights conveyed" ([17 U.S.C. 204](https://www.govinfo.gov/link/uscode/17/204?type=usc&year=mostrecent&link-type=html)). Look for an assignment or a broad licence clause in the integration contract. Machine builders often keep their code and lock it. Vendor function-block libraries stay the vendor's under the software licence.

Customer contracts are the other limit. A plant that runs a customer's process, as many contract manufacturers do, may hold process data that the supply agreement treats as the customer's. The [contract manufacturer guide](https://origindatapartners.com/guides/contract-manufacturer-design-data-ai) covers that case. This page is general information, not legal advice.


## Does licensing historian data expose our process recipes?

It can, if you export everything. Setpoints, ramp rates and recipe parameters sit in the same historian as the alarms. Decide which tags stay out, then license the rest under confidentiality terms.

Federal law protects information as a trade secret only when "the owner thereof has taken reasonable measures to keep such information secret" and it "derives independent economic value, actual or potential, from not being generally known" ([18 U.S.C. 1839](https://www.govinfo.gov/link/uscode/18/1839?type=usc&year=mostrecent&link-type=html)). A licence is a disclosure, so it needs confidentiality, access limits and a bar on publication and resale.

In practice plants use three levels. Recipe and setpoint tags for current products are withheld. Tags that are useful but sensitive are kept with values scaled or names replaced. Everything else, usually most of the alarm history and equipment health data, goes as it is. Old product lines and retired equipment are the easiest place to start.


## Do export controls apply to plant-floor data?

They can. Export rules cover technical information, not only goods, and sending data to a buyer abroad or to a foreign person in the United States can be an export. Classify the data before a sample leaves.

Under the Export Administration Regulations, technology means "Information necessary for the" development, production or use of an item ([15 CFR 772.1](https://www.ecfr.gov/current/title-15/section-772.1)). An export includes "An actual shipment or transmission out of the United States", and releasing technology or source code "to a foreign person in the United States" is a deemed export ([15 CFR 734.13](https://www.ecfr.gov/current/title-15/section-734.13)).

Defence work is stricter. Technical data under the arms regulations includes "Information, other than software as defined in § 120.40(g), which is required for the design, development, production, manufacture, assembly, operation, repair, testing, maintenance, or modification of defense articles" ([22 CFR 120.33](https://www.ecfr.gov/current/title-22/section-120.33)). Process data from a line that makes controlled parts needs an export review, not a guess.

Utilities have a further rule. The North American reliability standard on information protection exists "To prevent unauthorized access to BES Cyber System Information" ([NERC CIP-011-3](https://www.nerc.com/pa/Stand/Reliability%20Standards/CIP-011-3.pdf)). Network layouts, addresses and security settings of grid control systems do not belong in a dataset.


## Checklist: site and customer identifiers to remove

Plant data carries few personal details, but it is full of names that identify sites, customers and products. Work through the list before the sample, not after.

- Tag names and descriptions that contain customer, product or site names.
- Operator names, logins and badge numbers in alarm acknowledgements and audit trails.
- IP addresses, hostnames, node names and network diagrams.
- Passwords, licence keys and connection strings left in code and comments.
- Batch and lot IDs that match customer order numbers.
- Comments in code that name engineers, integrators or customers.
- HMI screenshots that show a logo, an address or a product name.
- Vendor library blocks you hold under licence and may not pass on.
- GPS positions and asset addresses for pipelines, wells and substations.

Keep a mapping table so that tag A-101 is the same tag in every file after renaming. General methods are in the [de-identification guide](https://origindatapartners.com/guides/de-identify-company-data-before-licensing).


## Example: one downtime event, exported as a linked record

A fictional bottling plant pulls one filler stop from a retired line. Nothing is written for the buyer. The work is exporting each piece and keeping the keys that join them.

| Piece | Record | Source system |
| --- | --- | --- |
| Trend before the stop | Thirty minutes of filler speed, bowl level and valve position at the logged rate | Historian export by tag and time range |
| Alarm sequence | First-out alarm, the four that followed, acknowledge and reset times | SCADA alarm journal |
| Code in service | The PLC project version running that day, with the interlock rung | Version control or the last backup before the date |
| Cause | Downtime reason code and the technician note | MES or downtime tracker |
| Fix | Work order: level sensor replaced, part number, labour time | Maintenance system |
| Outcome | Hours run before the next stop on the same alarm | Historian and downtime tracker |

Fictional example. Export timestamps in UTC with the time zone stated, keep units and tag descriptions, and state the sample rate and any compression. A few thousand events like this one describe a line better than a decade of bare trends.


## Common mistakes with plant-floor data

Most of them cut the links that give the data its value, or send something the plant did not own.

- Exporting hourly averages because the raw file was large.
- Sending time series without the tag descriptions and units.
- Counting integrator and machine-builder code as the plant's own.
- Including current recipes and setpoints in the first sample.
- Leaving passwords and IP addresses in PLC comments.
- Skipping export classification because "it is only sensor data".
- Wiping the historian server when a line is retired, before anything is exported.


## Plant-floor records as a data asset

One published company range applies (one buyer's public listings, read 15 November 2026). The listing for manufacturing and industrial companies runs from $100,000 to $1.5 million and says the figure "Depends on record types and history". Read it as a published ceiling, not a typical outcome: it assumes owned records, several years of history and no customer restriction.

A retired line is often the best first candidate, because its recipes are no longer live and its servers are about to be switched off. There is no standard timetable. The order is usually the same: a written description of the archive, a small sample, a rights review, agreed terms, then delivery. The rights review sets the pace. Makers of engineered products can read the [engineered product manufacturer guide](https://origindatapartners.com/guides/engineered-product-manufacturer-data), and licence terms are explained in the [data licensing agreement guide](https://origindatapartners.com/guides/data-licensing-agreement). Origin Data Partners is an independent data partner that helps companies get their data bought. A buyer may pay Origin when a deal closes.

Does your company hold records like these? [See what your data is worth](https://origindatapartners.com/check-fit).


## Questions and answers

### Can a plant license its historian data to AI companies?

Yes, when the plant owns the records and no customer contract restricts them. A public listing read on 15 November 2026 asks industrial operators for alarm logs and process data with measured results and several years of history. Recipes and setpoints for current products are usually withheld.

### Can you sell ladder logic or PLC code for AI training?

Only code you own. Code written by your own staff belongs to the company. Code written by an integrator or machine builder belongs to them unless a signed writing assigns it, and vendor library blocks stay under the vendor licence.

### Who owns PLC code that an integrator wrote for our plant?

The integrator, by default. Under 17 U.S.C. 101 commissioned software is not a work made for hire, and under 17 U.S.C. 204 a transfer of copyright needs a signed writing. Check the integration contract for an assignment or licence clause.

### Is raw sensor data worth anything to AI buyers?

Little by itself. Value comes from linking the trend to the alarm sequence, the cause, the fix and the result. Keep tag descriptions, units, batch context and the work order that followed each stop.

### Does licensing SCADA data expose process recipes?

Only if recipe and setpoint tags are exported. Plants usually withhold those tags, scale or rename sensitive ones and license the rest under confidentiality terms, which also helps keep trade secret protection under 18 U.S.C. 1839.

### How much historian history is enough for an AI data licence?

One public listing asks for several years of history. Length matters less than coverage: a set that includes start-ups, changeovers, upsets and repairs is more useful than a longer run of steady operation.

## Sources

- [NIST glossary: programmable logic controller (from SP 800-82r3)](https://csrc.nist.gov/glossary/term/programmable_logic_controller)
- [NIST glossary: supervisory control and data acquisition (from SP 800-82r3)](https://csrc.nist.gov/glossary/term/supervisory_control_and_data_acquisition)
- [NIST glossary: data historian (from SP 800-82r3)](https://csrc.nist.gov/glossary/term/data_historian)
- [IEC 61131-3:2013: programmable controllers, programming languages (abstract)](https://webstore.iec.ch/en/publication/4552)
- [17 U.S.C. 101: definitions (work made for hire)](https://www.govinfo.gov/link/uscode/17/101?type=usc&year=mostrecent&link-type=html)
- [17 U.S.C. 204: execution of transfers of copyright ownership](https://www.govinfo.gov/link/uscode/17/204?type=usc&year=mostrecent&link-type=html)
- [18 U.S.C. 1839: definitions (trade secret)](https://www.govinfo.gov/link/uscode/18/1839?type=usc&year=mostrecent&link-type=html)
- [15 CFR 772.1: definitions of terms (technology)](https://www.ecfr.gov/current/title-15/section-772.1)
- [15 CFR 734.13: export (Export Administration Regulations)](https://www.ecfr.gov/current/title-15/section-734.13)
- [22 CFR 120.33: technical data (International Traffic in Arms Regulations)](https://www.ecfr.gov/current/title-22/section-120.33)
- [NERC CIP-011-3: cyber security, information protection (PDF)](https://www.nerc.com/pa/Stand/Reliability%20Standards/CIP-011-3.pdf)

## Related reading

- [License Industrial Maintenance Records to AI Companies](https://origindatapartners.com/guides/license-industrial-maintenance-records-ai)
- [License Manufacturing Quality Records to AI Companies](https://origindatapartners.com/guides/license-manufacturing-quality-records-ai)
- [CAD Data and Quality Records for AI: Manufacturer Guide](https://origindatapartners.com/guides/engineered-product-manufacturer-data)
- [Who Owns Contract Manufacturer Tooling and Process Data?](https://origindatapartners.com/guides/contract-manufacturer-design-data-ai)
- [How to de-identify company data before licensing it](https://origindatapartners.com/guides/de-identify-company-data-before-licensing)
- [Data licensing agreement: what to include](https://origindatapartners.com/guides/data-licensing-agreement)
