It would be strange for a digital-safety business to pretend it does not use AI, so we will not. We do use it, in specific and limited ways, and we would rather write down how than leave you to guess. This is that page, and it is meant to be plain enough that you can hold us to it.

What it does is the dull, useful stuff. It helps us sort and summarise the flood of scam reports so we can spot a wave early. It drafts first versions of guidance that a person then rewrites. It helps us search our own records faster. In every one of these cases it is speeding up work that a human still owns, checks and signs off. It is a tool on the bench, not a colleague with a login to your accounts.

What it never does is decide alone anything that affects you. It does not choose whether your setup passes a check. It does not send you advice unread. It does not make a call about your data or your money without a named person having looked. Where AI touches a decision about a real person, a human is accountable for the outcome, and that is not negotiable for us.

We also think you are owed honesty about uncertainty. When something we tell you is a confident, checked fact, we say so. When it is a judgement or an estimate, we label it as such rather than dressing it up. AI is fluent, and fluency reads as confidence whether or not the thing is true. Pretending otherwise would be exactly the kind of quiet overreach we warn other people about.

This page will change as the tools do, and we will date it when it does. If any of it stops being true in practice, that is a fault worth telling us about. The point of writing it down is that a promise you can read is a promise you can check.