Is your business data used to train AI?
It is the first question most owners ask about AI, and the right one. The answer depends almost entirely on which terms the software runs under, and most people never check.
Consumer apps and business terms are different
The free chat apps you can download are consumer products. Their terms can allow your conversations to be used to improve future models, sometimes unless you find the setting that turns it off.
Software companies reach the same models through the providers' paid business services, under different terms. Under the business terms of established providers such as Anthropic and OpenAI, the data you send is not used to train their models by default, and it is kept only for a short, fixed period set out in those terms, typically for safety monitoring, before being deleted.
The practical risk for most businesses is not the provider. It is a member of staff pasting a customer list or a contract into a free app on their own account, because it was quicker. If that is happening in your business, and in most it is, it is the first thing to fix.
What we do
- We use leading AI models under business terms only.
- We do not use free consumer AI tools for client work.
- Each job reads only what it needs, not everything it could reach.
- Your data stays yours. If we stop working together, you take it with you in a usable form.
Five questions to ask any supplier
- Which AI providers do you use, and under which terms? "We use ChatGPT" is not an answer. Business service or consumer app?
- Is any of my data used to train a model? The answer should be a plain no, with the terms to show it.
- How long is my data kept, and where? Days or years? In the UK, the EU or elsewhere?
- Who can see it? Inside your company, inside theirs, and inside the AI provider.
- If we part company, how do I get my data back? In what format, and how quickly?
A supplier who cannot answer these in plain words is telling you something.
Where AI should not touch the figures
There is a second, less discussed reason for care. AI is very good at reading an email or a PDF. It is not reliable at arithmetic or at remembering exact figures across a long document. So in everything we build, totals and counts are worked out in code from the real system, never by the AI. The AI reads and drafts; the code counts; a person approves.
Have a job like this? Tell us in a few lines what gets copied, checked or chased by hand. We will come back with questions, not slides.
Book a callNext: What does it actually cost to automate part of a business?