Offline AI for PLC programming means running the language model on the engineer's own computer, so prompts, programs and drawings never leave the machine. A local model run through a general tool such as Ollama does this: once it is set up it needs no network at all, which is what many UAE solar, storage and substation sites require.
Cloud AI assistants are useful in an office. On an energy site they usually are not allowed. This article explains why, how a local-model set-up works in practice, the questions an OT security team will ask before it lets the laptop through the gate, and where the web tool LADX fits.
Why energy sites keep AI offline
A battery storage plant or a 132 kV substation is critical infrastructure. Its control network is separated from the internet on purpose, and the engineering laptop that connects to it is treated as a potential way in. Pasting a client's PLC program into a public chatbot sends that program to a third party, which most EPC contracts and most asset owners forbid.
The UAE is paying close attention to this. In February 2026 an OT Security First Forum was held in Abu Dhabi under the patronage of the UAE Cyber Security Council, focused on protecting the operational technology behind national energy and utilities infrastructure (Security MEA). The head of cyber security for the UAE government told the forum that IT and OT convergence has "dissolved traditional security perimeters".
The international reference points say the same thing. ISA/IEC 62443 treats security as a shared responsibility between asset owners, integrators and service providers, and Part 2-4 sets security programme requirements for service providers. NIST SP 800-82 Rev. 3, published in September 2023, is a practical guide to OT security that many UAE clients cite in their own policies.
How a local model works on an isolated laptop
The model runs on the laptop itself, served by a tool such as Ollama. The weights are several gigabytes, so you pull them on a connected machine and carry them across under the site's removable-media procedure.
- Install Ollama and pull a model on a connected machine, then move the laptop into the site's controlled environment.
- Check the files. Compare any published hashes for the installer and model files, and record the result.
- Switch automatic updates off. Bring new versions in through the same controlled route.
- Work in a project folder. Keep prompts, drafts and accepted code in a real directory you can back up, copy to an approved USB stick and hand over.
- Describe the logic. The model drafts rungs or Structured Text, and you compile it with the vendor's tools before anything else.
Why a small local model needs checking
A model that fits on a laptop is weaker than the largest cloud models, so treat every draft as unproven. Compile it, review it and simulate it. A compiler proves code is valid, not that the interlock is sufficient for the hazard. That judgement stays with the engineer, and on an energy site it also goes through the client's review and the protection engineer.
What is Ollama? Running Local LLMs Made Simple
IBM Technology
This video is hosted on YouTube. Playing it loads content from Google, which may set cookies.
Watch on YouTubeWhere LADX fits, and where it does not
LADX is web only and free. It has no desktop app, no offline mode and no installer, and it cannot use a local Ollama model, because its AI requests go from ladx.ai's server. That makes it unsuitable for client programs on an isolated energy site. "LADX Studio" is now just the name of the signed-in workspace at ladx.ai/studio.
It suits non-sensitive drafting, learning and checking on a connected machine. Chat needs a free account and runs on free models through LADX's shared account, rate limited per account, or on your own provider key for paid models. LADX never sells data or uses it to train AI models, but on the free tier the company running the free model may keep prompts and answers. Generated ladder is validated and given a seal-in check, Structured Text is compiled by the LADV compiler, and problems are listed beside the answer. Code that passes can be run on a Wartens LADV virtual PLC at ladx.ai/vplc, which is not real time, not safety rated and cannot control machinery.
What it does not do
- It never writes to a live controller. Neither the model nor LADX downloads to a PLC. You move tested code into the vendor's engineering tool and download it under the site's own procedure.
- It is not a security product. A local model removes one data-leak path. Laptop hardening, removable-media control and access management are still the site's job.
- It does not replace testing. Simulated logic still goes through FAT and SAT on the real system.
Questions your OT security team will ask
| Question | Answer for a local model on an isolated laptop |
|---|---|
| Does it make outbound connections? | No, once set up with updates off |
| Where do prompts and programs go? | To the local model on the same computer |
| Can we verify the files? | Yes, against published hashes where they exist, recorded in the software inventory |
| Does it connect to our PLCs? | No. Keep it in engineering, with downloads in the vendor tool |
| Can we use LADX instead? | Not for client programs: it is web only and its free tier may retain prompts and answers |
Where this fits with Wartens in the UAE
In the UAE, Wartens engineers SCADA and controller logic for solar-plus-storage and commercial sites, for EPC contractors and developers. Our energy systems page covers that work, and our guide to solar plus storage SCADA explains what EPCs need from the control system. For PLC work on existing plants, see automation engineering.
Next step: write down the data rules for the site, then test a local model on a non-production laptop with your security team. For learning and non-sensitive checking, try the tools at ladx.ai. If you need controls engineering for a UAE energy project delivered under your OT policy, tell us about the site through contact.
