What LADX AI is
LADX AI drafts and explains PLC programs, converts programs between platforms, drafts engineering documents such as test protocols and specifications, and answers questions about your project. It does this by sending your request to a large language model made by another company, together with the parts of your project it needs. Wartens Ltd is the provider of LADX AI as an AI system; the models themselves are made and run by others.
LADX AI is not a high-risk AI system under the EU AI Act. It is not a safety component of any machine and must never be used as one.
Which model answers you
- With your own key (OpenRouter, Anthropic, OpenAI or a compatible provider), LADX AI uses the provider and model you choose, under your own account.
- Without a key, LADX AI uses free models through LADX's shared OpenRouter account. It tries OpenRouter's own free router first and then other free models, and the line-up changes as providers add and withdraw free models. The answer tells you which model produced it.
Free models are slower and less precise on technical detail than the best paid models, and they are rate limited at busy times.
What is sent, and where
Each request carries your message, the earlier part of the conversation, and the project context needed to answer: for example the rungs and tags of the program open in front of you, or the sections of a document being drafted. It never carries your password or your provider keys to anyone except the provider the key belongs to.
Requests go to OpenRouter in the United States and on to the company running the model. OpenRouter says it does not log the content of prompts and answers by default. On the free tier, the company running the model may keep your request and use it to train its models. Do not send personal data, client confidential information or anything you are not allowed to share to the free tier. With your own key, what the provider keeps depends on your settings with that provider.
What LADX keeps
We keep your conversations, the code you accept, and anything you ask LADX AI to remember, so that you can come back to them; you can delete them, or your whole account, at any time. Our audit trail records that code was generated or accepted, and when, but not the code itself. We do not use your conversations, code or documents to train or fine-tune any AI model.
Checks before you see code
Generated Structured Text is checked automatically when it is shown, for structural mistakes such as blocks that are not closed and unbalanced brackets. On the web this is not a full compiler: it does not check data types or that every variable is declared. Where the full IEC 61131-3 compiler is installed, the code is also compiled. These checks catch some syntax errors. They cannot tell whether the logic does what your machine needs, and they say nothing about safety.
What it cannot do
- It can be wrong while sounding certain, including about instruction behaviour, addressing, timing and vendor differences.
- It does not know your machine, your risk assessment or your site standards unless you tell it, and even then it can misapply them.
- It must never be used for a safety function. Safety logic belongs in certified safety hardware and in programs developed under IEC 61508, IEC 62061 or ISO 13849.
Everything it produces is a draft for a competent engineer to review and test before it reaches equipment, as section 5 of the Terms of Service requires. You decide what is used; LADX AI only proposes.
How AI output is marked
Every AI panel says that you are working with an AI system. Documents LADX produces in Word, PDF and HTML carry a machine-readable mark in their properties stating that they were produced with LADX and may contain AI-generated content. Program code is not marked, as the EU AI Act does not require it for source code.
Tell us about a problem
If LADX AI produced something wrong in a way that could matter on a real machine, please tell us at hello@ladx.ai with the conversation or code. Reports like that are how it gets better.