AI at KUZUB

01Our approach

KUZUB LTD makes software for companies and specialists in design, construction and manufacturing. Some of our software uses artificial intelligence (AI), and we develop engineering models of our own. This page sets out the principles we follow wherever our products use AI. The terms of each product add the detail for that product, including whether the model-improvement use described below applies to it and whether it is on or off by default.

KUZUB LTD is a private limited company registered in Scotland with company number SC890997. Its registered office is at Suite 2/3, 48 West George Street, Glasgow G2 1BP, United Kingdom.

02People make the decisions

Our software is a tool for qualified people. With or without AI, it calculates, checks, drafts and proposes; it does not design, and it does not decide for anyone. Every result that someone relies on, whether a calculation, a selection, a schedule, an estimate, a drawing or a document, is a draft until a competent person has checked and accepted it. The professional responsibility for that result stays with that person and their organisation, and does not pass to KUZUB LTD.

AI can be wrong. Factual statements and figures in AI output must be checked before anyone relies on them, like every other draft our software produces.

03Other companies' AI models and your data

This section applies to Caldaia Studio, and to each of our other products that uses other companies' AI models.

We use third-party AI models under commercial terms that do not allow the provider to train its models on the data we send. Providers may keep inputs and outputs for a limited period to detect abuse, as set out in the Sub-processors list.

We design each feature to send a provider only the data that the feature needs, and we use other companies' models only as their terms allow. We do not train our own models on the output of other companies' AI models where their terms do not allow it.

04Our own engineering models

We develop our own engineering models, not general-purpose chat models, to help our products calculate, check and plan. The following applies in Caldaia Studio. Where another of our products uses project data in this way, its terms say whether model improvement applies to it and what the default is, and those terms apply instead.

We improve our own engineering models - not general-purpose chat models - using de-identified technical structure taken from projects: which catalogue items are used, how they are connected and laid out, pipe sizes, lengths, materials and insulation, work types and quantities, currency, month and country. Names, addresses, prices you set, logos, notes and anything else that identifies you, your organisation or your clients are removed before use; each item keeps only a coded reference to your account and organisation, which we use solely to honour your choices and never to identify you. For personal accounts this is on by default and you can switch it off at any time in Account settings; for company workspaces it is off unless an administrator switches it on. If you switch it off, or delete your account, no further data from your projects is used, and data already taken from your projects is deleted within 30 days and left out of any model we train after that; models trained before then are not changed, and nothing else in the service changes. We do not sell this data and we do not give it to AI providers to train their models.

05Transparency

We tell people when they are interacting with an AI system, unless that is obvious from the context, and we mark AI-generated content wherever the law requires it, including in machine-readable form. We keep these notices short and factual.

06Law and oversight

Before an AI feature is offered to customers, we assess it against the law that applies to it, including the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) and UK and EU data protection law. Personal data used with AI is processed as the privacy notice of each product describes, with the same rights as any other personal data.

We do not use AI to take decisions about people that have legal or similarly significant effects on them.

07Security and control

Our own models run only in environments that we operate, or that a customer operates where it chooses to run our software itself. The vulnerability disclosure policy of each product covers its AI features too, including any way to make a feature reveal data it should not, or act outside its purpose; for this website, that policy is at kuzub.co.uk/security.

08Contact

Write to hello@kuzub.co.uk with questions about AI in our products, to privacy@kuzub.co.uk about personal data, and to security@kuzub.co.uk about security issues.

09Changes to this page

We update this page when our approach to AI changes. The date at the top shows when it last changed. This is version 1, the first version published by KUZUB LTD.